Report

Renewable acrylonitrile production

See allHide authors and affiliations

Science  08 Dec 2017:
Vol. 358, Issue 6368, pp. 1307-1310
DOI: 10.1126/science.aan1059

A sweet source to make acrylonitrile

Much of the attention directed toward displacing petroleum feedstocks with biomass has focused on fuels. However, there are also numerous opportunities in commodity chemical production. One such candidate is acrylonitrile, a precursor to a wide variety of plastics and fibers that is currently derived from propylene. Karp et al. efficiently manufactured this compound from an ester (ethyl 3-hydroxypropanoate) that can be sourced renewably from sugars. The process relies on inexpensive titania as a catalyst and avoids the side production of cyanide that accompanies propylene oxidation.

Science, this issue p. 1307

Abstract

Acrylonitrile (ACN) is a petroleum-derived compound used in resins, polymers, acrylics, and carbon fiber. We present a process for renewable ACN production using 3-hydroxypropionic acid (3-HP), which can be produced microbially from sugars. The process achieves ACN molar yields exceeding 90% from ethyl 3-hydroxypropanoate (ethyl 3-HP) via dehydration and nitrilation with ammonia over an inexpensive titanium dioxide solid acid catalyst. We further describe an integrated process modeled at scale that is based on this chemistry and achieves near-quantitative ACN yields (98 ± 2%) from ethyl acrylate. This endothermic approach eliminates runaway reaction hazards and achieves higher yields than the standard propylene ammoxidation process. Avoidance of hydrogen cyanide as a by-product also improves process safety and mitigates product handling requirements.

Acrylonitrile (ACN) is one of the most widely used monomers in the chemical industry, with more than 14 billion pounds produced annually for use in plastics, rubbers, resins, acrylic fibers, and polyacrylonitrile (PAN)–based carbon fibers (1, 2). The market outlook for PAN-derived carbon fibers in particular is projected to grow by 11 to 18% annually, driven by interest in reducing the weight of vehicles and aircraft (2). Today, industrial ACN production is conducted via the Sohio process, which converts propylene to ACN via ammoxidation over a bismuth molybdate–based catalyst. First-generation catalysts for the process were developed in the 1950s, achieving ~55% molar ACN yields from propylene (3). This discovery spurred decades of research to improve ACN yields (4), resulting in state-of-the-art materials capable of producing ACN at molar yields of 83% from propylene (5). However, fluctuations in the price of the propylene feedstock translate directly to ACN price volatility. The carbon fiber industry is especially sensitive to these ACN price fluctuations, because roughly 2 lbs of ACN are required to generate 1 lb of fiber (6). Thus, concerns about propylene price volatility have motivated the search for alternative approaches to propylene ammoxidation to produce ACN. Substantial efforts have focused on the ammoxidation of propane, which is a cheaper substrate than propylene and has a lower carbon footprint, but propane is still derived from fossil-based sources (7). Environmentally sustainable routes to ACN have been described from renewable feedstocks such as glycerol (810) and glutamic acid (11); these efforts are summarized in (2). The most promising method to date is glycerol dehydration to acrolein followed by ammoxidation to ACN, achieving yields of ~60% (8). Accordingly, there is a clear need to develop sustainable, cost-effective bio-based ACN manufacturing routes.

To that end, we present a route from ethyl 3-hydroxypropanoate (ethyl 3-HP, derived from microbially produced 3-hydroxypropionic acid, 3-HP) to ACN at molar yields of ≥90%. This approach originates from reports of carboxylic acids being converted to nitriles when passed over solid acids with ammonia (12). However, this reaction is made difficult by the relatively low volatility of carboxylic acids and the corrosiveness of their vapors on equipment. Esters are less corrosive, more volatile, and in general more stable than their acid counterparts; using esters as substrates with this chemistry could enable a more viable route to renewable ACN.

To test the viability of converting ethyl 3-HP to ACN, we first conducted a steady-state temperature scan by passing ethyl 3-HP over TiO2 with an 8:1 molar excess of ammonia (Fig. 1A and fig. S1). Ethyl 3-HP was consumed in conjunction with the appearance of ethyl acrylate as the temperature was increased from 150°C to 230°C. Increasing the temperature further, from 230°C to 320°C, produced ACN at the expense of ethyl acrylate. From this result, we posit that three sequential reactions occur (Fig. 1B) to form ACN. First, the primary alcohol undergoes dehydration to form ethyl acrylate and water, then ethyl acrylate undergoes aminolysis (Fig. 1B, reaction 2) to form acrylamide and ethanol, and finally acrylamide is dehydrated to produce ACN and water (Fig. 1B, reaction 3).

Fig. 1 Catalytic scheme for ethyl 3-HP dehydration and nitrilation to produce ACN.

(A) Steady-state yields of relevant reaction products produced when passing ethyl 3-HP over TiO2 as a function of reactor bed temperature. Complete reaction conditions and data set are provided in fig. S1. (B) The three reactions that are proposed to yield the results in (A). (C) Proposed mechanism from DFT calculations for the aminolysis of ethyl acrylate to form acrylamide and gaseous ethanol [reaction 2 in (B)]. (D) Proposed mechanism from DFT calculations for dehydration of adsorbed acrylamide to release gaseous acrylonitrile and water [reaction 3 in (B)].

The primary alcohol of 3-HP (or ethyl 3-HP) is known to readily dehydrate to an acrylate (13). Ester aminolysis to an amide has been reported using wet chemistry techniques (14), and amide dehydration to nitriles is also known (15). We refer to these latter reactions together as nitrilation (Fig. 1B, reactions 2 and 3). Little work has been published on nitrilation overall, especially in systems where ester and ammonia vapors are passed over heterogeneous catalysts to form nitriles (1618).

Given sparse mechanistic information for this chemistry, we performed periodic density functional theory (DFT) calculations to probe the catalytic mechanism over a TiO2(101) surface. Alcohol dehydration was predicted to proceed via an E2 mechanism (fig. S2), similar to well-known dehydration reactions (19). For the aminolysis of ethyl acrylate, DFT results suggested a stepwise mechanism catalyzed by partial dissociation of NH3 on TiO2. H-NH2 bond scission exhibits the highest barrier, +111 kJ/mol, in this mechanism (Fig. 1C). DFT results for the final reaction suggest that acrylamide undergoes dehydration to form ACN via surface-mediated tautomerization to its enol form followed by dehydroxylation (Fig. 1D). The highest-barrier step in the nitrilation reaction was predicted to be the deprotonation of the N-H group, with a barrier of +101 kJ/mol. See figs. S3 to S17 and tables S1 to S5 for details and energetics of these pathways.

Kinetic measurements performed at low conversions (figs. S18 and S19) revealed apparent activation energies consistent with the energetics of the rate-limiting steps from DFT. For the ester dehydration, an apparent activation energy of +100 ± 4 kJ/mol was determined from low-conversion experiments of ethyl 3-HP over TiO2 (fig. S18). The measured activation energy compares favorably to the +112 kJ/mol calculated for dehydration (figs. S2 and S7). The apparent activation energy of nitrilation was measured by performing low-conversion experiments with ethyl acrylate and ammonia over TiO2 (fig. S19). Here, an apparent activation energy of +103 ± 12 kJ/mol also compares favorably to the +111 kJ/mol barrier for the H-NH2 bond scission from DFT (figs. S11 and S12), which is the highest predicted nitrilation barrier.

Results of total-conversion experiments performed in a tandem bed reactor are shown in Fig. 2 and fig. S20. In the first reactor, ethyl 3-HP was dehydrated over TiO2 to form ethyl acrylate and water in quantitative yield at 260°C. A molar excess of 2:1 ethanol to ethyl 3-HP was used as the feed to the first reactor to suppress acrylic acid formation (figs. S20 and S21). The product vapors from the first reactor were then mixed with ammonia and passed over a second bed of TiO2 at 315°C to form ACN, ethanol, and water. This operation achieved ACN yields of 90 to 92% for ~12 hours on stream with minimal deactivation. The overall heat of reaction is calculated to be endothermic by +203 kJ/mol (fig. S22).

Fig. 2 Catalytic conversion of synthetic and biologically derived ethyl 3-HP to ACN.

Total-conversion reactions of ethyl 3-HP in tandem catalytic beds demonstrate 90 to 92% yields of ACN. The percent yield of ACN is shown with the carbon balance for the reaction. See fig. S20 for the complete data set for all observed reaction products and the reaction conditions used. Approximately every 12 hours, the reaction was stopped and the catalyst regenerated. The rightmost graph represents data collected using ethyl 3-HP separated from a 3-HP cultivation on glucose using an engineered E. coli strain, showing performance identical to that of the synthetic ethyl 3-HP substrate. As illustrated in the reactor schematic, glass beads were packed in the headspace of the reactors to achieve uniform gas mixing over the catalyst.

At the 12-hour time point for each run, the reaction was stopped and the catalyst removed and regenerated in air at 550°C. Images of fresh, spent, and regenerated catalyst, as well as data from pyridine diffuse reflectance infrared Fourier transform spectroscopy, Brunauer-Emmett-Teller isotherms, x-ray diffraction, and acid site density measurements (figs. S23 to S26 and table S6), indicate that the regeneration cycle restores the measured characteristics to those of the fresh sample. The regenerated catalyst showed identical performance to the fresh sample (Fig. 2). Thermogravimetric analysis with Fourier transform infrared spectroscopy (TGA-FTIR) measurements of the gas released during catalyst regeneration showed that NOx was not produced (fig. S27), thereby abating the need for exhaust cleanup during regeneration.

To ascertain whether this chemistry exhibits different behavior on a biologically derived substrate, we produced 3-HP via glucose cultivation using an engineered Escherichia coli strain (20) (fig. S28). The cultivation used fed-batch dissolve oxygen–based control to feed glucose to the bioreactor, and resulted in a titer of 25.8 g/liter (supplementary text and figs. S29 and S30). After glucose cultivation, ethyl 3-HP was separated and recovered from the broth (figs. S31 and S32), yielding 97% purity. The ethyl 3-HP was catalytically processed (Fig. 2) and achieved performance identical to that of synthetic ethyl 3-HP.

The high yields of the nitrilation chemistry to produce ACN from microbially derived ethyl 3-HP (Fig. 2) allow us to propose a potential industrial-scale process for the hybrid biological and catalytic transformation of lignocellulosic sugars to ACN. This process (Fig. 3) exhibits several notable modifications from the process demonstrated at bench scale. First, 3-HP production ideally would be conducted at low pH [below the pKa of 3-HP (21)]. The advantage gained from low-pH cultivation is that the acidification step is no longer required during separations, and neutralization is not required during cultivation, thus avoiding generation of waste salt. This would improve the process economics and sustainability (22). Low-pH strains to produce 3-HP at industrially relevant titers, rates, and yields are under development (21, 23). The second difference is that dewatering would occur at scale using a simulated moving bed (SMB) where 3-HP is adsorbed to a resin and eluted off with ethanol. Results in fig. S33 indicate that polybenzimidazole (PBI) works well for this, allowing 3-HP to be completely recovered (24). Third, ethyl acrylate, instead of ethyl 3-HP, would be separated via reactive distillation. Reactive distillation combines esterification, alcohol dehydration (Fig. 1B, reaction 1), and product separation into a single unit operation. This is accomplished using the same industrial process previously developed for the production of acrylate esters from β-propiolactone (BPL) (25). In that process, BPL is added to an ethanol solution acidified with sulfuric acid in a continuous stirred tank reactor, where ring opening occurs to form 3-HP and then esterifies and dehydrates to form the acrylate ester (25). The solution is then fed to a distillation column where the acrylate ester is recovered and purified in the column overhead (25, 26); see figs. S34 and S35 for details. The process could be stopped here and ethyl acrylate sold as the product at a predicted selling price of $0.48/lb, which may be attractive given that the current market price of ethyl acrylate is $0.79/lb. However, the market for ACN is approximately 10 times as large, and the potential for lowering carbon fiber prices has much greater societal impact to lower the greenhouse gas footprint of automobile transportation by reducing vehicle weight (27, 28).

Fig. 3 Conceptual process diagram for renewable ACN production from biomass sugars.

Process depiction of the unit operations proposed to produce ACN from lignocellulosic sugars via a hybrid biological-catalytic upgrading approach.

Recognizing that ethyl acrylate (rather than ethyl 3-HP) would be fed to the nitrilation unit in the conceptual process shown in Fig. 3, we performed total-conversion experiments with ethyl acrylate and ammonia over TiO2 in a single catalytic bed (Fig. 4 and fig. S36). These runs resulted in a 98 ± 2% maximum yield of ACN with ethyl acrylate as a substrate. The ACN yield from ethyl acrylate is higher than that from ethyl 3-HP because of decreased carbon deposition on the catalyst, which is likely attributable to the reduced presence of water when using ethyl acrylate as the substrate. The carbon balance shown in Fig. 4 is slightly above 100% because of slightly decreased ethyl acrylate conversion after 12 hours on stream, leading to a slight buildup of ethyl acrylate (fig. S36). A small amount of ethyl acrylate present in the reactor outlet leads to measurement uncertainty because of a low signal-to-noise ratio. The overall heat of ethyl acrylate nitrilation is calculated to be endothermic by +81 kJ/mol (fig. S37).

Fig. 4 Catalytic conversion of ethyl acrylate to ACN.

Total-conversion reactions of ethyl acrylate in a single catalytic bed, demonstrating maximum ACN yields of 98 ± 2%. The percent yield of ACN is shown with the carbon balance for the reaction. See fig. S36 for the complete data set for all observed reaction products and the reaction conditions used. Approximately every 12 hours, the syringe was refilled (as shown) and every 18 hours the reaction was stopped and the catalyst removed and regenerated.

In a scaled process (Fig. 3), it would be economically beneficial to operate the nitrilation reactor in a pure ammonia atmosphere without the use of N2 as a diluent. Our online analytical system precludes measurements in a pure ammonia atmosphere (see supplementary text), but catalyst deactivation studies (figs. S38 to S40) performed using ethyl acrylate as a substrate allowed calculation of a regeneration cycle time needed under more concentrated conditions (fig. S41 and supplementary text). These calculations estimate that under full-scale conditions (310°C, 0.89 atm NH3, 0.11 atm ethyl acrylate), the entire catalytic bed must be regenerated every 30 s to maintain ACN yields above 98%. Thus, the fourth modification in the full-scale process model is that the recovered ethyl acrylate is fed to a riser reactor with continuous catalyst regeneration (Fig. 3 and fig. S42). This continuously regenerates the catalyst with a cycle time of 30 s. The liberated alcohol is recovered downstream and recycled (figs. S43 and S44). A simplified schematic of the modeled process is shown in Fig. 3; see figs. S34, S35, and S42 to S44 for details of the unit operations.

On the basis of this proposed process, we performed a techno-economic analysis predicting the selling price of ACN at $0.89/lb from lignocellulosic sugars and $0.76/lb from sucrose. These target prices are in the range of fossil fuel–derived ACN prices between $0.40 and $1.00/lb over the past decade (29). Additionally, the greenhouse gas emissions from this process were estimated to achieve a 14.1% improvement relative to propylene-derived ACN. In addition to the sugar platform, other low-value feedstocks including glycerol (30, 31) and waste gases are being pursued for the production of 3-HP. If realized, these platforms could further lower the ACN selling price when coupled to nitrilation. See tables S7 to 15 and figs. S34, S35, and S42 to S46 for details of these analyses, a discussion of the parameters used, and model sensitivity to these parameters.

Beyond ACN, nitrilation may have broader applications by providing a facile link from carboxylic acid or ester production to nitriles. Several biologically derived carboxylates are now being operated at industrial scale (e.g., succinic, lactic, itaconic, and fumaric acid) (32, 33) and the nitrile derivatives of these acids could likely be readily obtained via nitrilation. An economic advantage also may exist in coupling nitrilation to bioprocesses, because separation of the ester can often be more economical than separating the free acid (34, 35).

For ACN production, nitrilation provides a number of green chemistry benefits over propylene ammoxidation: (i) Near-quantitative yields of ACN can be obtained from this reaction, whereas state-of-the-art ammoxidation catalysts achieve ~80 to 83% ACN yield (5). (ii) The reaction is endothermic (figs. S22 and S37) and does not require O2, enabling facile process control. By comparison, ammoxidation is highly exothermic, requiring specialized reactors to avoid runaway reactions (36). (iii) Unlike ammoxidation, nitrilation does not produce hydrogen cyanide, mitigating toxicity and handling requirements. (iv) The cost of TiO2 is approximately 30% that of ammoxidation catalysts (4, 37). (v) The process provides a cost-comparable, sustainable route to ACN with potential greenhouse gas emission offsets from a renewable feedstock.

Because the Sohio process depends on propylene, ACN prices have historically been tied to petroleum prices and therefore volatile. Industrial deployment of an ACN production process using an alternative feedstock, such as described in this work, could stabilize the ACN price by unhinging it from sole dependence on fossil resources. Combined with further process development, the use of nitrilation could lead to a sustainable process for bio-based ACN, and ultimately to products such as renewable carbon fiber.

Supplementary Materials

www.sciencemag.org/content/358/6368/1307/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S48

Tables S1 to S15

References (38100)

References and Notes

  1. Acknowledgments: Supported by U.S. Department of Energy (DOE) Bioenergy Technologies Office grant DE-FOA-0000996. Computer time was provided by Extreme Science and Engineering Discovery Environment (XSEDE) allocation MCB-090159 at the Texas Advanced Computing Center and by the National Renewable Energy Laboratory Computational Sciences Center, supported by the DOE Office of Energy Efficiency and Renewable Energy under contract DE-AC36-08GO28308. We thank L. Berstis, S. Kim, and P. Kostestkyy for helpful discussions regarding the catalytic mechanism; D. Salvachúa and X. Wang regarding bioreactor cultivation; B. Black, K. Ramirez, and M. Reed for analytical assistance; and members of the Renewable Carbon Fiber Consortium for helpful discussions. E.M.K., T.R.E., D.R.V., and G.T.B. are inventors on patent application WO 2017/143124 A1, US 2017/018272 submitted by the Alliance for Sustainable Energy that covers nitrile production from bio-based feedstocks. R.T.G. is a co-inventor on intellectual property related to biological 3-HPA production, which is now owned by Cargill. All data generated in this study are in the supplementary materials. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes.
View Abstract

Navigate This Article