You are on page 1of 92

www.sciencemag.

org/content/358/6368/1307/suppl/DC1

Supplementary Materials for

Renewable acrylonitrile production

Eric M. Karp, Todd R. Eaton, Violeta Sànchez i Nogué, Vassili Vorotnikov, Mary J. Biddy,
Eric C. D. Tan, David G. Brandner, Robin M. Cywar, Rongming Liu, Lorenz P. Manker,
William E. Michener, Michelle Gilhespy, Zinovia Skoufa, Michael J. Watson,
O. Stanley Fruchey, Derek R. Vardon, Ryan T. Gill, Adam D. Bratis, Gregg T. Beckham*

*Corresponding author. Email: gregg.beckham@nrel.gov

Published 8 December 2017, Science 358, 1307 (2017)


DOI: 10.1126/science.aan1059

This PDF file includes:


Materials and Methods
Supplementary Text
Figs. S1 to S48
Tables S1 to S15
References
TABLE OF CONTENTS:
MATERIALS AND METHODS ..................................................................................................... 7
CATALYSTS .................................................................................................................................................... 7
CATALYST CHARACTERIZATION ........................................................................................................................ 7
REAGENTS FOR NITRILATION EXPERIMENTS ...................................................................................................... 8
CATALYTIC REACTOR EXPERIMENTS ................................................................................................................ 8
GAS PHASE FTIR SYSTEM............................................................................................................................... 8
YIELD AND CARBON BALANCE CALCULATIONS FOR FIGURES IN MAIN TEXT ........................................................... 9
COMPUTATIONAL METHODS............................................................................................................................. 9
SYNTHESIS OF 3-HP STANDARD .................................................................................................................... 10
METABOLIC ENGINEERING OF ESCHERICHIA COLI............................................................................................ 10
MEDIA AND CULTIVATION CONDITIONS ............................................................................................................ 10
DO-STAT FED-BATCH PRODUCTION OF 3-HP .................................................................................................. 11
METABOLITE ANALYSIS ................................................................................................................................. 11
SEPARATION / RECOVERY OF ETHYL 3-HP FROM FERMENTATION BROTH .......................................................... 12
GC METHOD FOR ETHYL 3-HP PURITY ANALYSIS ............................................................................................ 12
PBI RESIN TESTING ...................................................................................................................................... 12
PROCESS DESIGN DETAILS ............................................................................................................................ 13
ECONOMIC ANALYSIS APPROACH .................................................................................................................. 15
ECONOMIC ANALYSIS RESULTS ..................................................................................................................... 15
SENSITIVITY ANALYSIS .................................................................................................................................. 15
LIFE CYCLE ASSESSMENT .............................................................................................................................. 16
SUPPLEMENTARY TEXT.......................................................................................................... 17
FURTHER DISCUSSION OF COMPUTATIONAL RESULTS ...................................................................................... 17
FURTHER DISCUSSION OF SCALED UP PROCESS MODEL USED IN THE ECONOMIC ANALYSIS ................................ 18
SUPPLEMENTARY FIGURES ................................................................................................... 20

S1
Figures:
Figure S1. Complete reaction conditions and dataset (with molar units) for observed reaction products for the data
presented in Figure 1A of the main text. ....................................................................................................................... 20
Figure S2. Reaction mechanisms predicted from DFT calculations for the dehydration of ethyl 3-HP to form gaseous
ethyl acrylate and water (Reaction (1) in Figure 1B). ................................................................................................... 21
Figure S3. Profile (left) and top (right) view of anatase TiO2(101) model used in DFT modeling. The layer relaxed in all
optimizations (atoms closest to the surface) is shown as spheres; the layers kept in their optimized bulk positions are
represented with a wireframe. ...................................................................................................................................... 22
Figure S4. Convergence tests performed in this work: (A) TiO2(101) slab energy as a function of k-point mesh, (B)
NH3 adsorption energy difference as a function of cutoff energy, and (C) NH3 adsorption energy difference with the
number of TiO2(101) bulk layers. The inset images, from left to right, represent one, three, and six bulk TiO2(101)
layers, respectively. ...................................................................................................................................................... 23
Figure S5. Ethyl-3-hydroxypropanoate (E3HP) dehydration mechanisms over TiO2: (A) dissociative pathway and (B)
concerted E2 pathway. Purple arrows represent the electron flow for the reactions following the intermediate. The
reactions associated with the dissociative pathway are E3HP adsorption (D1), O-H scission (D4), C-O and C-H
scission to produce ethyl acrylate (D7) and a series of H2O formation steps not shown here. The E2 pathway
proceeds through the same adsorption, followed by simultaneous C-OH and C-H scission to form ethyl acrylate (D8),
followed by H2O formation (D6). ................................................................................................................................... 24
Figure S6. Key intermediates in ethyl-3-HP dehydration over TiO2(101). Surface atoms are shown as a wireframe;
surface Ti and O (Os) atoms involved in bonding are shown as spheres for clarity. Bond distances are presented in Å.
...................................................................................................................................................................................... 25
Figure S7. Gibbs free energy diagram for ethyl-3-HP dehydration over TiO2(101) via an E2 mechanism at 310°C.
Free energy levels (kJ/mol) are shown in blue. Transition states are shown in the inset images: simultaneous C-OH
scission and deprotonation to form ethyl acrylate (left) and water formation (right). Bond lengths (Å) are shown in
black. ............................................................................................................................................................................ 26
Figure S8. Ethyl acrylate (EA) aminolysis mechanisms considered in this work: (A) concerted, single-step pathway,
(B) stepwise pathway, and (C) surface acrylate pathway. The concerted pathway involves NH3 and EA coadsorption,
followed by a single-step ethanol and acrylamide formation (A5). The stepwise pathway involves EA conformational
change (A12), partial NH3 decomposition (A8), and subsequent NH2 insertion (A6) to form a stable 1-amino-1-ethoxy-
2-propen-1-ol surface intermediate (AEP*). This intermediate undergoes rapid C-O scission (A9) to form acrylamide,
followed by ethanol formation (A11). The third pathway involves direct C-O scission (A10) to form a surface acrylate
intermediate, partial NH3 decomposition (A8), and subsequent NH2 insertion (A7). .................................................... 27
Figure S9. Key intermediates in ethyl acrylate aminolysis over TiO2(101). Surface atoms are shown as a wireframe;
surface Ti and O (Os) atoms involved in bonding are shown as grey and dark red spheres, respectively, for clarity.
Bond distances are presented in Å. ............................................................................................................................. 28
Figure S10. Gibbs free energy diagram for ethyl acrylate aminolysis over TiO2(101) via the concerted mechanism at
310°C. Free energy levels (kJ/mol) are shown in blue. Transition state for the simultaneous H-NH2 scission, C-O
scission, NH2 insertion and O-H formation is shown in the inset images. Bond lengths (Å) are included in black. ..... 29
Figure S11. Gibbs free energy diagram for ethyl acrylate aminolysis over TiO2(101) via the stepwise mechanism at
310°C. Free energy levels (kJ/mol), which are referenced to clean TiO2(101) and gas-phase ethyl acrylate (EA) and
NH3, are shown in blue. The first two barrier-less steps represent the sequential adsorption of NH3 and EA (A2 and
A1’), respectively. Transition states are shown in the inset images, from left to right: ethyl acrylate conformational
change (A12), H-NH2 scission (A8), NH2 insertion (A6), C-O and O-H scission (A9), and O-H formation (A11); bond
lengths (Å) are included in black. The last two barrier-less steps are EtOH and acrylamide desorption (A4’ and A3),
respectively................................................................................................................................................................... 30
Figure S12. Gibbs free energy diagram for ethyl acrylate aminolysis over TiO2(101) via the acrylate mechanism at
310°C. Free energy levels (kJ/mol), which are referenced to clean TiO2(101) and gas-phase ethyl acrylate (EA) and
NH3, are shown in blue. The first two barrier-less steps represent the sequential adsorption of NH3 and EA (A2 and
A1’’), respectively. The next step (-58 to 35 kJ/mol) is a conformational change/diffusion of EA (A13). Transition
states are shown in the inset images for the next four reactions, from left to right: C-O scission (A10), H-NH2 scission
(A8), O-H formation (A11), and NH2 insertion (A7); bond lengths (Å) are included in black. The last two barrier-less
steps are EtOH and acrylamide desorption (A4’ and A3), respectively. ....................................................................... 31

S2
Figure S13. Acrylamide dehydration mechanisms considered in this work: surface-mediated tautomerization to
acrylamide enol form followed by (A) sequential N-H and C-OH scission or (B) simultaneous N-H and C-OH scission,
(C) ethanol-forming single-step tautomerization followed by N-H and C-OH scission and (D) dissociation of
acrylamide without tautomerization via a series of N-H and C-O scission steps. Surface-mediated tautomerization for
paths A and B involve N-H scission (AD5) followed by O-H formation (AD13). Path A further involves N-H scission
(AD6) and C-OH scission (AD11). Path B follows a simultaneous N-H and C-OH scission (AD8). Path C involves
hydrogen transfer to simultaneously form ethanol and tautomerize acrylamide to its enol form (AD10). The co-
adsorbed surface ethanol is not shown along this path. Path D involves direct C-O scission (AD9). .......................... 32
Figure S14. Key intermediates in acrylamide dehydration over TiO2(101). Surface atoms are shown as a wireframe;
surface Ti and O (Os) atoms involved in bonding are shown as grey and dark red spheres, respectively, for clarity.
Bond distances are presented in Å. ............................................................................................................................. 33
Figure S15. Gibbs free energy diagram for acrylamide dehydration over TiO2(101) via surface-mediated
tautomerization and stepwise N-H and C-OH scission at 310°C. Free energy levels (kJ/mol), which are referenced to
clean TiO2(101) and gas-phase acrylamide (AAm), are shown in blue. The first barrier-less steps represent AAm
adsorption (AD1) and conformational change (AD4), respectively. Transition states are shown in the inset images,
from left to right: N-H scission (AD5), O-H formation (AD13), second N-H scission (AD6), C-OH scission (AD11) and
H2O formation (AD15); bond lengths (Å) are included in black. The last barrier-less step is H2O desorption (AD3). .. 34
Figure S16. Gibbs free energy diagram for acrylamide dehydration over TiO2(101) via surface-mediated
tautomerization and a single-step N-H and C-OH scission at 310°C. Free energy levels (kJ/mol), which are
referenced to clean TiO2(101) and gas-phase acrylamide (AAm), are shown in blue. The first barrier-less steps
represent AAm adsorption (AD1) and conformational change (AD4), respectively. Transition states are shown in the
inset images, from left to right: N-H scission (AD5), O-H formation (AD13), a concerted N-H and C-OH scission (AD8)
and H2O formation (AD15); bond lengths (Å) are included in black. The last barrier-less step is H2O desorption (AD3).
...................................................................................................................................................................................... 35
Figure S17. Gibbs free energy diagram for acrylamide dehydration over TiO2(101) via direct H-transfer
tautomerization followed by step-wise N-H and C-OH scission at 310°C. Free energy levels (kJ/mol), which are
referenced to acrylamide (AAm) coadsorbed with ehtoxy (EtO) and H species TiO2(101), are shown in blue.
Transition states are shown in the inset images. The first involves AAm tautomerization to its enol form while
simultaneous forming ethanol (AD10). The next transition states, from left to right, are: N-H scission (AD6), C-OH
scission (AD11) and H2O formation (AD15). Bond lengths (Å) are included in black. The last barrier-less step is H2O
desorption (AD3). ......................................................................................................................................................... 36
Figure S18. Arrhenius plot for the dehydration of the 3 position –OH group in ethyl 3-HP to form ethyl acrylate and
water. Substrate conversion is <15%. .......................................................................................................................... 37
Figure S19. Arrhenius plot for the nitrilation of ethyl acrylate to form ethanol, water and acrylonitrile. This analysis
-1
yields an apparent activation energy for the nitrilation of ethyl acrylate of 103 ± 12 kJ mol . Substrate conversion is
<15%. ........................................................................................................................................................................... 38
Figure S20. Complete dataset and conditions used for the results shown in Figure 2 of the main text. The graph
displays all reaction products measure in molar units. Complete reaction conditions used are given in the grey box. 39
Figure S21. High-conversion experiments were performed to maximize the yield of ethyl acrylate formed when
dehydrating the primary alcohol on ethyl 3-HP. Here ethyl 3-HP was passed over the TiO2 catalyst at 260°C with a
2:1 molar excess of ethanol. The excess ethanol was needed to suppress the equilibrium formation of acrylic acid,
see reaction shown in the bottom right. In the results shown in the graph, 100% yield of ethyl acrylate was obtained
with 100% mass balance using the parameters listed in the grey box. ........................................................................ 40
Figure S22. Reaction enthalpy for the dehydration and nitrilation of ethyl 3-HP to form acrylonitrile water, and ethanol.
The reaction enthalpy was calculated from the values in Table S2. (+179.7+-234+2*-241.8)-(-45.9+-695.1) = +203
kJ/mol. Note here that the heat of formation of ethyl 3-HP is not available so in its place we use the heat of formation
of ethyl lactate, a structural isomer. .............................................................................................................................. 41
Figure S23. Fresh, spent and regenerated TiO2 catalyst used for the experiments shown in Figure 2. ...................... 42
Figure S24. DRIFTS spectrum of pyridine on fresh (green) and regenerated (blue) TiO2............................................ 43
Figure S25. Nitrogen physisorption of fresh and spent TiO2 with BET surface areas indicated in the legend. ............ 44
Figure S26. X-ray diffractogram of fresh, spent, and regenerated TiO2. ...................................................................... 45

S3
Figure S27. TGA-FTIR results during coke burnoff of the spent catalyst shown in Figure S23. Note here no NOx was
observed. Only CO2 and water was observed. ............................................................................................................. 46
Figure S28. Metabolic pathway for 3-HP production in recombinant E. coli BGHP strain. ........................................... 47
Figure S29. Fed-batch DO-stat cultivation of the E. coli BGHP strain using glucose as a carbon source. .................. 48
Figure S30. Fed-batch DO-stat cultivation of the E. coli BGHP strain using DDAP-EH corn stover hydrolysate. ........ 49
Figure S31. Cartoon depiction of the nine-step procedure used for separating ethyl 3-HP from fermentation broth. The
process is described in detail in materials and methods section. The recovered ethyl 3-HP was converted to
acrylonitrile as described in the main text and in Figure 2. ........................................................................................... 50
Figure S32. Pictures of the glucose fed fermentation broth as it was processed through the nine-step procedure
depicted in Figure S31. ................................................................................................................................................. 51
Figure S33. (A) Loading curve for acidified 3-HP broth from glucose finding a loading capacity of 230 mg/g dry resin
for the polybenzimidazole (PBI) resin. Note that a capacity of at least 50 mg acid / g dry resin is needed for the resin
to be industrially relevant (136). (B) Elution curve of 3-HP from the resin using ethanol as the eluent. ....................... 52
Figure S34. Graphical representation of the process flow diagram for the fermentation operation to produce 3-HP at
low pH and the dewatering operation used to produce a concentrated stream of 3-HP in ethanol via a simulated
moving bed (SMB) and distillation column. .................................................................................................................. 53
Figure S35. Graphical representation of the reactive distillation operation used in the economic analysis. Design
parameters were taken from previously published work available in references (62, 63). .......................................... 54
Figure S36. Complete dataset and conditions used for the results shown in Figure 4 of the main text. The graph
displays all reaction products measure in molar units. Complete reaction conditions used are given in the grey box. 55
Figure S37. Reaction enthalpy for the nitrilation of ethyl acrylate to form acrylonitrile ethanol and water. This reaction
enthalpy was calculated from the values in .................................................................................................................. 56
Figure S38. Activity loss curves (A/A0) displaying the deactivation of the titania catalyst as a function of time at 310°C.
Four curves are present representing different partial pressures of ammonia while keeping the partial pressure of
ethyl acrylate constant at 1.08 kPa. These results indicate that increasing the partial pressure of ammonia decreases
the rate of catalyst deactivation. The fitted lines represent an empirical fit using both this data set and that shown in
Figure S46 discussed in the SI text. ............................................................................................................................. 57
Figure S39. Activity loss curves (A/A0) displaying the deactivation of the titania catalyst as a function of time at 310°C.
Four curves are present representing different partial pressures of ethyl acrylate while keeping the partial pressure of
ammonia constant at 4.23 kPa. These results indicate that increasing the partial pressure of ethyl acrylate increases
the rate of catalyst deactivation. The fitted lines represent an empirical fit using both this data set and that shown in
Figure S45 discussed in the SI text. ............................................................................................................................. 58
Figure S40: Correlation between the grams of coke burned off of the TiO2 catalyst as a function of activity lost,
assuming a linear fit. .................................................................................................................................................... 59
Figure S41: Activity loss curve calculated from the empirical model for activity loss (Equation S5) for the full-scale
process design equations. This result calculates that a 0.5 minute regeneration cycle time is needed for the catalyst
to maintain ≥ 98% yields of ACN from ethyl acrylate.................................................................................................... 60
Figure S42. Graphical representation of the process flow diagram used to model the nitrilation reaction unit
operations. As described in the SI, nitrilation is modeled to occur in a riser reactor system, where catalyst
regeneration occurs continuously. Reaction products are captured in a DMSO gas stripping tower that contains 100
ppm of 4-hydroxy TEMPO as a polymerization inhibitor solved in the DMS. The DMSO is then passed through a
distillation column to recover the water, ACN, ethanol, and the DMSO and inhibitor is recycled back to the stripping
tower. ............................................................................................................................................................................ 61
Figure S43. Graphical representation of the process flow diagram used to model the first section of the ACN
separation train where ACN is recovered at a purity of 97% from a distillation tower and decanter system. The
bottoms product from the column is passed to the section of the ACN separation train shown in Figure S44. ........... 62
Figure S44. Graphical representation of the process flow diagram used to model the second section of the ACN
separation train. Here the bottoms product from the previous distillation column (Figure S43) is fed to the middle of a
second distillation column to recover ethanol which is dried using molecular sieves and the ethanol product is
recycled back to the SMB dewatering operation in Figure S34. ................................................................................... 63

S4
Figure S45. Cost breakdown for the major process areas in the production of ACN from lignocellulosic biomass. .... 64
Figure S46. Tornado plot summarizing the results of the sensitivity analysis performed on the process. ................... 65
Figure S47. Life cycle stages considered in the FTG pathway for biomass-derived ACN from 2G sugars.................. 66
Figure S48. Life cycle stages considered in the 1G sugar-derived ACN. ..................................................................... 67

S5
Tables:
Table S1. Anatase TiO2 computed and experimental lattice parameters and volume. ................................................ 68
Table S2. Ethyl acrylate nitrilation enthalpies of formation and reaction enthalpies in kJ/mol. .................................... 69
Table S3. Reaction energetics associated with ethyl-3-hydroxypropanoate dehydration in kJ/mol (D) at 0 K. The
quantities ΔE, EA,f, and EA,r represent the reaction energy and the barriers for the forward and the reverse reaction,
respectively. All values (ZPE-corrected in parentheses) are given in kJ/mol. Abbreviations stand for ethyl-3-
hydroxypropanoate (ethyl 3-HP), ethyl acrylate (EA), and an alkoxide intermediate resulting from ethyl 3-HP
dehydrogenation (ethyl 3-HP-H). Star symbol (*) denotes a surface-bound species. .................................................. 70
Table S4. Reaction energetics for aminolysis (A) at 0 K. The quantities ΔE, EA,f, and EA,r represent the reaction
energy and the barriers for the forward and the reverse reaction, respectively. All values (ZPE-corrected in
parentheses) are given in kJ/mol. Abbreviations stand for ethyl acrylate (EA), ethanol (EtOH), an ethoxide
intermediate resulting from EtOH O-H scission (EtO), 1-amino-1-ethoxy-2-propen-1-ol intermediate (AEP), acrylamide
(AAm), and surface acrylate (Acr). The numbering for EA and EtOH surface species (e.g., EAn) represents the
various conformations along the respective paths. Star symbol (*) denotes a surface-bound species. ....................... 71
Table S5. Reaction energetics for acrylamide dehydration (AD) at 0 K in kJ/mol. The quantities ΔE, EA,f, and EA,r
represent the reaction energy and the barriers for the forward and the reverse reaction, respectively. All values (ZPE-
corrected in parentheses) are given in kJ/mol. Abbreviations stand for acrylamide (AAm), enol form of acrylamide
(Enol), ethanol (EtOH), an ethoxide intermediate resulting from EtOH O-H scission (EtO), partially dehydrogenated
acrylamide resulting from a single N-H scission (AAm-H) and two N-H scission reactions (AAm-2H), partially
dehydrogenated enol form of acrylamide resulting from N-H scission (Enol-H) and acrylonitrile (ACN). The numbering
for AAm surface species (e.g., AAm1) represents the conformations along the respective reaction paths. Star symbol
(*) denotes a surface-bound species. ........................................................................................................................... 72
Table S6: Measured acid site density and BET surface area for the fresh, spent, and regenerated TiO2 catalyst used
in the reactions shown in Figure 2 of the main text ...................................................................................................... 73
Table S7. Delivered feedstock composition assumed in the process design. .............................................................. 74
Table S8. Summary of key process parameters and yields utilized in the baseline process design............................ 75
Table S9. Capital cost breakout for process design. .................................................................................................... 76
Table S10. Financial assumptions and design basis. .................................................................................................. 77
Table S11. Summary of economics for integrated ACN production design using lignocellulosic biomass. ................. 78
Table S12. Cost of raw materials. ................................................................................................................................ 79
Table S13. Key inventory data reported per kg of bio-ACN produced from lignocellulosic sugars. ............................. 80
Table S14: Key inventory data reported per kg of bio-ACN produced from sugars from Number 11 commodity (raw)
sugar (1G) .................................................................................................................................................................... 81
Table S15: Acrylonitrile production life cycle GHGs comparison.................................................................................. 82

S6
MATERIALS AND METHODS
Catalysts
TiO2 was obtained from Johnson Matthey. The physical form of the TiO2 catalyst was 0.5 mm diameter
spheres, and it was used as received without any further treatment. Samples will be made available upon
request.

Catalyst characterization
Catalyst characterization techniques were applied to both fresh and spent catalyst samples to determine
physicochemical properties before and after reactions. In this section, “fresh” catalyst refers to a catalyst
that has been characterized as-received from the supplier, while “spent” catalyst refers to catalyst that
has been on stream for ethyl acrylate nitrilation for >12 hours.
The total number of acids sites was measured with ammonia temperature programmed desorption (TPD)
on an Altamira Instruments AMI-390 system. Catalyst samples (~100 to 200 mg) were packed into a
-1
quartz tube and heated to 600°C at 10°C min in 10% O2/Ar flowing at 25 sccm and held for 1 hour to
pretreat the catalyst. The samples were then cooled to 120°C, flushed with 25 sccm He for 10 minutes,
then saturated with ammonia by flowing 25 sccm of 10% NH3/He over the samples for 30 minutes at
120°C. Excess ammonia was removed by flushing with 25 sccm He for 10 minutes. The samples were
-1
then heated to 600°C in 25 sccm He at 30°C min , holding at 600°C for 30 minutes, and the desorbed
ammonia was measured with a thermal conductivity detector (TCD) that monitored the catalyst bed
effluent. The TCD was calibrated after each experiment by introducing 7 pulses of 10% NH3/He from a 5
mL sample loop into a stream of 25 sccm He. Acid site quantification was performed assuming an
adsorption stoichiometry of one ammonia molecule per site (38). This technique was applied to both fresh
TiO2 and spent TiO2 to determine the change of acid site density before and after reaction. The acid site
-1 -1
density for fresh TiO2 was 160 μmol g , while the acid site density for spent TiO2 was 200 μmol g .
The relative amount of Lewis to Brønsted acid sites was determined by pyridine adsorption diffuse-
reflectance fourier-transformed infrared spectroscopy (pyridine DRIFTS), using a Thermo Nicolet iS50 FT-
-1
IR spectrometer operating at 4 cm resolution equipped a Harrick Praying Mantis controlled-environment
chamber and KBr windows. Fresh catalyst samples (~50 mg) were loaded into the chamber and pre-
-1
treated by heating in 100 sccm N2 with a ramp rate of 10°C min to 300°C and held at this temperature
for 1 hour, then cooled to 150°C. A background spectrum was then collected of the clean catalyst surface
before pyridine vapor was introduced by bubbling 100 sccm N2 through liquid pyridine and through the
-1
catalyst bed for 10 minutes. The pyridine-saturated surface was then heated at 10°C min and held at
300°C for 30 minutes to remove pyridine that was not chemically bound to the surface, and then cooled to
150°C. A spectrum was then collected of the pyridine-modified catalyst and the absorption bands near
-1 -1
1445 cm (Lewis) and 1545 cm (Brønsted) and the relative absorption coefficients of these features
(ɛB/ɛL = 0.76), combined with total acid site density from ammonia TPD, were used to determine the
number of Brønsted and Lewis acid sites (39, 40). Pyridine DRIFTS of fresh and regenerated TiO2 (Figure
S24) showed that only Lewis acid sites were present, with adsorbed pyridine bands due to Lewis acid
-1 -1
sites appearing at 1443, 1574, 1603, and 1612 cm (41-43) and a band at 1490 cm attributed to
physisorbed pyridine (41).
Nitrogen physisorption measurements to determine surface area were performed on a QuadraSorb SI
instrument, using the BET method. Samples were dehydrated at 200°C under vacuum prior to analysis.
2 -1 2 -1
The isotherms are shown in Figure S25 and show surface areas for fresh (49 m g ), spent (27 m g ),
2 -1
and regenerated (50 m g ) TiO2.
To quantify carbon laydown on spent TiO2, and to identify the chemical identity of the carbon laydown,
thermogravimetric analysis with Fourier transform infrared spectroscopy (TGA-FTIR) was performed on a
ThermoScientific TGA-FTIR system, equipped with a Nicolet 6700 FTIR. Spent TiO2 was loaded into the
-1
system and heated under flowing zero air (50 sccm) at a rate of 20°C min to 600°C. Gas-phase
combustion products from the catalyst surface were tracked with the FTIR. The major combustion
products were CO2 and H2O, while small amounts of CO and CH4 were also detected. The catalyst mass

S7
loss and H2O and CO2 signals are shown as a function of temperature in Figure S27. No nitrogen-
containing products were observed with the FTIR.
X-ray diffraction was performed on fresh, spent, and regenerated TiO2 to determine the crystal phase,
using a Rigaku Ultima IV diffractometer with Cu Kɑ radiation at a step size of 0.02°. Both anatase and
rutile phases were observed in all samples, with slightly varying ratios each crystal phase, as shown in
Figure S26.

Reagents for nitrilation experiments


Ethyl 3-hydroxypropanoate (>97 %) (obtained from Combi-Blocks), ethanol (Pharmco-AAPER, 200 proof,
lot# C16B08002), ethyl acrylate (Sigma-Aldrich lot # SHBC8912V), and acrylonitrile (Sigma-Aldrich,
lot#SHBF8717V), were used to calibrate the online FTIR detector, described below. Ethyl acrylate was
used as an inexpensive substrate to probe optimal reaction conditions given that ethyl 3-
hydroxypropanoate is much more costly and dehydrates rapidly to produce ethyl acrylate (
Figure S1 and Figure 1A in the main text). Research grade purity of anhydrous ammonia gas (lot#
9004407323O1) obtained from Mattheson TRIGAS and metered into the reactor system via a MKS mass
flow controller.

Catalytic reactor experiments


Reactions were performed in a custom-built flow reactor system with tandem catalytic beds. The
substrate (ethyl 3-HP) was injected into a heated line at 150°C and vaporized into an N2 carrier gas. The
flowrate of the N2 gas was controlled with a MKS mass flow controller calibrated for N2 and ranged from 0-
2,000 sccm. Concentrations of the ester reagent were always <5% vol/vol of the gas entering into the
reactor. Approximately 1 inch before the ester vapors entered the nitrilation reactor, ammonia gas was
blended into the gas stream using an MKS mass flow controller calibrated for NH3 gas and ranged from 0-
500 sccm. NH3 concentrations in the inlet gasses to the reactor were always <15% vol/vol. The reactors
used were 50 ml in volume with an ID of 0.8” and 6” of length. A K-type thermocouple was placed in the
center of the bed to measure and control the temperature of the reactor. For high conversion studies, the
reactor was fully packed with catalyst; generally this was ~45 g of catalyst using 0.5 mm diameter TiO2
spheres from Johnson-Matthey. The activation energy of ethyl 3-HP dehydration to ethyl acrylate (EA)
was measured over TiO2 without NH3 present in the gas phase. The conditions for these measurements
were: 2 to 5 g of TiO2, 2000 sccm N2, 0.077 mL min-1 ethyl 3-HP. Exhaust gasses from the reactor bed
were passed through the FTIR cell and then bubbled through chilled dimethylformamide (DMF) (4°C),
which contained ~100 ppm of 4-hydroxy-TEMPO as a polymerization inhibitor, held in a knockout pot.
Two knockout pots are present on the system and a valve positioned downstream of the FTIR system
allowed the exhaust gasses to be directed to the desired knockout pot. During an experiment, the exhaust
gasses were directed to one knockout pot until steady state was reached then the valve was switched to
direct the exhaust gasses into the second knockout pot. This allowed collection of non-volatile
components at steady-state conditions. The DMF solution in the knockout pot collected acrylonitrile
(ACN), ethyl acrylate, water, and ethanol from the reaction. Excess ammonia was not absorbed into the
DMF solution due to the aprotic nature of DMF. Custom written software in OPTO22 allowed manual or
automated control of the entire system.

Gas phase FTIR system


The reactor exhaust gas was monitored in real time using an MKS FTIR system (model 2030) with a 2 cm
path length gas cell. The gas cell was heated and maintained at 191°C. Calibration curves for ammonia,
ethanol, and water were produced by metering in liquid compounds with a Series I HPLC pump (Scientific
Systems Inc.), ethyl 3-hydroxypropanoate, ethyl acrylate, and ACN, were metered into the gas stream
using a syringe pump (New Era Pump Systems Inc. model NE 1000) into nitrogen carrier gas in heated
gas lines held at 150°C. Ammonia gas was blended into the N2 carrier gas using a separate mass flow
controller. Concentrations were varied between 1-18% vol/vol in N2 to generate calibration curves. The
spectra were appropriately fenced for quantification using the MKS software. The full method is available
upon request. Mass and carbon balances >90% were routinely achieved from the FTIR data for each
reaction reported in this work when steady state was reached.

S8
Yield and carbon balance calculations for figures in main text
The percent yield values reported in the main text Figure 1A, Figure 2, and Figure 3B were calculated via
Equation (S1):

!!"#,!"#$%#
𝑌!"# = × 100 (S1)
!!"#,!!!"#!$%&'(

Here ṅACN, outlet is the measured molar flowrate of ACN at the outlet of the reactor using the FTIR
instrument described above, and ṅACN, theoretical is the maximum theoretical molar flowrate of ACN at the
outlet of the reactor using the stoichiometry of Scheme 1 in the main text. For example, in Figure 2 of the
-1
main text, the molar flowrate of ethyl 3-HP into the reactor is 6.87E-4 mol min (determined by the
measured mass loss of the syringe pump over the time period of injection) thus ṅACN, theoretical is also 6.87E-
-1
4 mol min as a result of the 1:1 stoichiometry between ethyl 3-HP to ACN shown in Scheme 1 of the
main text.

Carbon balances reported in Figure 1A, Figure 2, and Figure 3B were calculated in Equation (S2):

!∗!!"#!, !"#$%# ! !∗!!"#,!"#$%# ! !∗!!"!!" !!!",!"#$%#


𝐶𝑎𝑟𝑏𝑜𝑛 𝑏𝑎𝑙𝑎𝑛𝑐𝑒 = × 100 (S2)
!!"#$%&, !!!"#!$%&'(

Here the measured flowrates of ethanol (ṅetoh, outlet), ACN (ṅACN, outlet), and ethyl 3-HP (ṅethyl 3-HP, outlet) were
multiplied each by their respective number of carbon atoms and summed together in the numerator of
equation (S2). The numerator represents the total measured molar flowrate of carbon at the outlet of the
reactor. The denominator of equation (2) is the theoretical total molar flowrate of carbon at the reactor
outlet determined from the known inlet molar flowrate of ethyl 3-HP (or ethyl acrylate in Figure 3 of the
main text) multiplied by 5.
Note that we opt to report the carbon balance instead of total mass balance because the excess amount
of ammonia used in these reactions (e.g., a 8:1 molar ratio for NH3 to ethyl acrylate was used) results in
the mass balance being dominated by the amount of ammonia in the reactor exhaust and thus mass
balances were always very close to 100%. Therefore, the carbon balance is a more accurate
representation of the reaction products (ACN, ethanol, and any unreacted substrate).

Computational methods
Periodic DFT calculations were carried out using Vienna Ab Initio Simulation Package (VASP), version
5.4.1 (44-47). The ion-electron interactions were described using the PAW potentials (48, 49). The
electron-electron exchange and correlation energies were computed using Perdew, Burke, and Ernzerhof
(PBE) functional (50). The van der Waals (vdW) forces were calculated using the method of Tkatchenko
and Scheffler (TS) as implemented in VASP (51). The DFT+U method was used to correct for on-site
Coulomb interactions by setting effective U parameter to 2.5 eV (52, 53). A 17x17x7 Γ-centered k-mesh
was used for bulk anatase TiO2 calculations. The Murnaghan-Birch equation of state was used to
estimate bulk parameters, which were also confirmed to within 0.001 Å via single-point energy
calculations (54, 55). The resulting lattice parameters, summarized in Table S1, are in excellent
agreement with previous computational and experimental findings (53, 56). Gas-phase optimizations were
performed in a 20×21×22 Å box using the Gaussian smearing method until all forces were lower than
-1
0.01 eV Å .
We used a 2x3 surface that is 3 bulk TiO2 layers deep to represent TiO2(101), shown in Figure S3. The
vacuum layer between surfaces was set to 30 Å to minimize periodicity effects. The bottom layers were
fixed in their bulk positions while the top layer was allowed to relax. A 425 eV cutoff energy and a 3x3x1
k-point mesh were chosen for all calculations. The relaxation was performed until all forces were lower
-1
than 0.02 eV Å using a conjugate-gradient algorithm. Dipole corrections were included in all calculations

S9
(57, 58). As shown in Figure S4, these parameter choices result in relative energy errors of less than 0.01
-1
eV (1 kJ mol ).
Adsorbate vibrational frequencies were calculated using the finite difference method with a displacement
of 0.015 Å. In these calculations, surface atoms were fixed and only adsorbate vibrations were assumed
to have non-cancelling contributions to Gibbs free energy. Even with rather strict optimization criteria,
some imaginary frequencies were unavoidable. In these cases, these frequencies were replaced with 25
-1
cm . We used the climbing nudged elastic band (cNEB) (59) method to locate the transition states. Each
transition state was found to have a single imaginary frequency and was confirmed to relax to initial and
final states by perturbing the transition state along the vibration associated with that imaginary frequency.
Higher-accuracy gas-phase calculations were carried out using G4-level via Gaussian 09 package
(revision A.2) (60) to obtain enthalpies of formation and heats of reaction (61).

Synthesis of 3-HP standard


3-hydroxypropionic acid can be purchased as a 30 wt.% solution in water from several chemical
manufacturers. However, these products are not sufficiently pure to provide accurate calibration curves
for analytical measurements. Therefore, a synthetic standard of sodium 3-hydroxypropanoate was
synthesized by adding 25 g of ß-propiolactone (Alfa Aesar) to 500 mL of ultra pure water. We note that ß-
propiolactone is very hazardous and a known carcinogen regulated by the Occupational Safety and
Health Administration (OSHA) and care must be taken by researchers who wish to replicate this
procedure. In our laboratory, a bottle of ß-propiolactone was opened inside a nitrogen filled glove bag
within a fume hood and added to the ultra pure water inside the glove bag. A spray bottle was then used
to wet the inside of the glove bag to ensure any vapors of ß-propiolactone were contacted with water,
ensuring ring opening of the lactone. The solution was removed from the glove bag and stirred overnight
to ensure complete ring opening of the lactone to form 3-hydroxypropionic acid in water. The solution was
then titrated with a 1 M solution of sodium hydroxide to its equivalency point (pH = 9.33) forming sodium
3-hydroxypropanoate, which was then dried down using rotoevaporation and the recovered salt was dried
in a 40°C vacuum oven for 12 hours. We note that the sodium salt of 3-HP is stable; however the acid
form is not stable above ~30 wt.% in water owing to its tendency to self esterify. For analytical
measurements, the salt was used to produce calibration curves of known concentration and when
contacted by the sulfuric acid mobile phase during HPLC analysis, the salt form of 3-HP was transformed
to the acid form before reaching the detector. See section titled “metabolite analysis” below for details on
the HPLC method used.

Metabolic engineering of Escherichia coli


E. coli BG strain, kindly provided by Dr. Michael D. Lynch (Department of Biomedical Engineering &
Chemistry, Duke University), was generated as follows: the genes ackA, pta, poxB, ldhA, adhE, and pflB
(encoding acetate kinase, phosphate acetyltransferase, pyruvate oxidase, D-lactate dehydrogenase,
aldehyde-alcohol dehydrogenase, and pyruvate formate-lyase, respectively) were deleted from strain
BW25113 (Coli Genetic Stock Center, CGSC#: 7636) (62) and a temperature sensitive fabI allele, inactive
at 37℃, was used to decrease the metabolic flux into the fatty acid synthesis pathway and increase the
accumulation of malonyl-coA (63). Genes of the 3-HPA pathway, mcr from Chloroflexus aurantiacus
encoding a malonyl-CoA reductase (GenBank: AAS20429.1) and ydfG from Escherichia coli K12
(ecogene: EG12345), were cloned under the control of the constitutive J23119 promoter into a high copy
vector pSMART (Lucigen Corporation). The resulting plasmid pSMART-HCkan-J23119-Camcr-ydfG was
cloned into E. coli BG following manufacturer’s instructions (CloneSmart Cloning Kit, Lucigen Corporation)
Kit, Lucigen Corporation) and the strain was named BGHP (Figure S28).

Media and cultivation conditions


-1 -1 -1
LB medium consisted of 10 g L tryptone, 5 g L yeast extract, 5 g L NaCl. FM5 medium contained 2.63
-1 -1 -1 -1 -1
g L K2HPO4, 1.38 g L KHPO4, 3 g L (NH4)2SO4, 2.19 g L citric acid x H2O, 5 g L yeast extract, 0.82
-1 -1 -1 -1
g L MgSO4 x 7H2O, 20 mg L Fe2(SO4)3 x 7H20, 1.2 mg L CoSO4 x 7H20, 10 mg L CuSO4 x 5H20, 0.6
-1 -1 -1
mg L ZnSO4 x 7H20, 0.4 mg L Na2MoO4 x 2H2O, and 0.2 mg L H3BO3 (Sigma-Aldrich).

S10
Corn stover biomass, provided by Idaho National Laboratory, was hammer-milled and filtered through a
rejection screen before de-acetylation with 0.4% (w/w) (64) NaOH solution to achieve a 8% (w/w) total
solids (TS) loading. After de-acetylation, pretreatment was carried out continuously using a horizontal
screw fed reactor (Metso Inc, Norcross, GA, USA) with dilute 1.0% H2SO4 (w/w) at 50% (w/w) TS. For the
enzymatic hydrolysis, Novozymes Cellic® CTec2 was added after adjusting the post pretreatment slurry
to 20% (w/w) TS with water and the pH to 5.2 using a 50% NaOH solution. Enzymatic hydrolysis was
carried out with slight agitation at 48°C for 168 h while maintaining a pH 5.2. The de-acetylated, dilute-
acid pretreated, enzymatically hydrolyzed (DDAP-EH) hydrolysate (64) was concentrated under vacuum
at ~55˚C for 3 days and stored at 4°C prior to use. The monomeric sugar composition of the resulting
-1 -1 -1 -1
concentrated hydrolysate was: 240.0 g L glucose, 137.5 g L xylose, 4.5 g L galactose, and 14.0 g L
-1 -1
arabinose. Feed for the glucose cultivation contained 500 g L glucose supplemented with 25 g L yeast
-1
extract. Concentrated DDAP-EH hydrolysate supplemented with 25 g L yeast extract was used as feed
in the hydrolysate cultivation.
Cell concentrations were determined from absorbance measurements at 600 nm on sample diluted to
TM
give an optical density (OD) below 0.4 (Genesys 20, ThermoFisher Scientific, Walthman, MA, USA).
Seed cultures were generated by inoculation of 1 L baffled shake flasks containing 100 mL of LB with
glycerol stock of the E. coli BGHP strain. Cells were aerobically grown in an orbital incubator (Innova
4330, New Brunswick, Eppendorf, Hauppauge, NY, USA) at 30°C, and 225 rpm for 16 h. Cells were
harvested by centrifugation and resuspended with 0.9% NaCl solution. The seed culture was used to
inoculate the bioreactors (Bioflo 310, New Brunswick) at initial OD of 0.1.

DO-stat fed-batch production of 3-HP


Fermentations were carried out in a DO (Dissolved Oxygen)-stat fed-batch mode. Aerobic conditions were
obtained by continuously sparging air at 1 vvm and pH was kept constant at 7.0 by addition of 15%
(NH4)OH. During the batch phase, the DO level was maintained above 25% by increasing the agitation
speed. In the glucose cultivation, medium FM5 supplemented with 10 g L-1 glucose was used for the
batch phase. When glucose was depleted, pulses of glucose feed corresponding to 1 mM glucose were
added to the bioreactor via DO-stat control mode, where a DO level of 65% was set as a trigger. In the
hydrolysate cultivation, concentrated DDAP-EH hydrolysate diluted to a final monomeric sugar
-1 -1 -1 -1
concentration of 5 g L glucose, 3.0 g L xylose, 0.1 g L galactose, and 0.3 g L arabinose, and
-1
supplemented with 5 g L yeast extract was used for the batch phase. When sugars were depleted
(confirmed by a sharp increase of DO level), pulses of hydrolysate feed corresponding to 0.45 mM
monomeric sugars were added to the bioreactor via DO-stat control, where a DO level of 65% was set as
a trigger. Every 24 hr, biotin and NaHCO3 were added to the bioreactor at a final concentration of 40 μg L-
1 and 20 mM, respectively. In all fermentations, temperature was switched from 30°C to 37°C once the
culture reached an OD of 20. Antifoam 204 (Sigma-Aldrich) was added in the feed at a final concentration
-1
of 3 mL L . Every 24 h, antibiotics Kanamycin (Sigma-Aldrich) and Gentamicin (Sigma-Aldrich) were
-1 -1
added to a final concentration of 50 μg L and 20 μg L , respectively.

Metabolite analysis
Cells were quickly separated by centrifugation; the supernant was filtered through 0.20 μm nylon
membrane filter (Whatman) and stored at 4°C until analysis. 3-HP, lactic acid, formic acid, acetic acid,
and ethanol analysis was performed on an Agilent 1100 LC system equipped with a G1362A refractive
index detector (Agilent Technologies). Each sample was injected at a volume of 20 μL onto an Aminex
HPX-87H 7.8 x 300 mm i.d., 9 µm column (BioRad) at an oven temperature of 55 °C with an isocratic flow
-1
of 0.01 N H2SO4 at 0.6 mL min . Analysis of glucose, galactose, xylose, and arabinose was performed
using an ICS-5000+ system consisting of an AS-AP autosampler, and a pulsed electrochemical detector
with a gold electrode and an Ag/AgCl reference electrode (Dionex Corporation). Samples were diluted to
a quantifiable range and 10 μL was injected on to a CarboPac SA-10 Dionex carbohydrates column (4 x
250 mm) paired with a CarboPac SA-10 guard column (4 x 50mm). Sugars were separated with an
-1
isocratic flow of 1 mM potassium hydroxide at 1.5 mL min prior to 5 min at 45°C. Following the sugar
separation a ramp program was used with increasing potassium hydroxide concentration and then an
equilibrium for a total run time of 15 min. Sugar standards of glucose (99.5% purity), galactose (≥ 99%

S11
purity), xylose (≥ 99% purity), and arabinose (≥ 98% purity) used to construct calibration curves between
-1
the range of 0.5 – 60 mg L were purchased from Absolute Standards Inc.

Separation / recovery of ethyl 3-HP from fermentation broth


Separation and recovery of ethyl 3-HP was achieved through the nine-step procedure outlined in Figure
S31. Fermentation was conducted, as described above, with the pH maintained with the addition of
ammonium hydroxide, producing the ammonium 3-hydroxypropanoate salt. The broth was first treated in
Step 2 of Figure S31 by removing cells and debris through centrifugation and filtration through a 0.2 μm
polyethersulfone filter (ThemoFisher Scientific). Protein was then removed from the cell free broth by
filtration through a 10 kDa filter (GE Part # UFP-10-C-3X2MA). In Step 3, activated carbon (3 w/v % for 4
hours) was then used to remove color bodies from the broth. The spent carbon was filtered off from the
broth and the clarified broth dewatered in Step 4 through roto-evaporation leaving a crust of ammonium 3-
hydroxypropanoate and sugar syrup. Next, in Step 5, the ammonium 3-hydroxypropanoate salt was
dissolved into ~1 L of ethanol through gentle heating (~50°C) of the dried material in ethanol. Note that
the ammonium 3-HP salt is soluble in ethanol, as is ammonium lactate, the structural isomer of
ammonium 3-HP (65), and the sugar syrup was insoluble in ethanol (66). This dissolution process allowed
the separation of the sugar syrup from the ammonium 3-HP salt via decanting. In step 6 the ethanol
solution containing dissolved ammonium 3-HP salt was acidified with the addition of a stoichiometric
amount of H2SO4. This produced ammonium sulfate, which is precipitated out of the ethanol solution and
was filtered off, and left 3-HP dissolved in ethanol. In Step 7, a catalytic amount of H2SO4 (0.025 mol
-1
H2SO4 mol 3-HP) was added to the 3-HP/ethanol solution and esterified under reflux for 100 minutes.
After 100 minutes of refluxing, the esterification solution was neutralized with the addition of 30 mL of ultra
high purity water saturated with sodium carbonate. Salts produced from neutralization were filtered off,
and the ethyl 3-HP product was isolated in step 8 through vacuum distillation. The product was then dried
over sodium sulfate. Purity of the ethyl 3-HP product was determined to be ~88% based upon the GC
chromatogram (method described below) on an area basis compared to the purchased synthetic standard
from Combi-Blocks. To increase the purity further, Dry Column Vacuum Chromatography (DCVC) (67)
was used in Step 9. Here ~30 g of the dried ester product was dissolved in dichloromethane and loaded
onto Celite (Alfa Aesar, lot #N13B029) through roto-evaporation. The loaded Celite (1 cm bed height) was
packed on top of a 7 cm bed of silica gel (Merck, Silica gel 60, lot #TA1817911323). The ester was eluted
through the column starting with pure pentane in the first fraction and increasing by 2 wt.% acetone in the
subsequent fractions until a final fraction of 20 wt.% acetone/80 wt.% pentane was reached. The volume
of each eluent fraction was 150 mL. The ester product was observed through testing each fraction on thin
layer chromatography (TLC) plates using a potassium permanganate stain. The pure fractions were
mixed together and the pentane/acetone solvent was blown off of the product. Final purity of the ester
product was determined to be 97.3% by GC area analysis compared to the purchased ethyl-3HP
standard. Pictures of the broth and product as it was processed in each separation step are shown in
Figure S32.

GC method for ethyl 3-HP purity analysis


GC analysis was performed on an Agilent HP-5MS column in an Agilent 6890N GC. Each injection was 1
-1
μL using a splitless He inlet at 280°C and 11.32 psi with a total flow of 7.6 mL min . The initial oven
-1
temperature was set to 35°C and held for 3 minutes before ramping at 15°C min to 150°C which was
-1
held for 1 minute. The system was then purged by ramping at 30°C min to 250°C where it was held for 5
-1
minutes. The Flame Ionization Detector (FID) was operated at 280°C with 35 mL min of hydrogen, 300
-1 -1
mL min of air, and 1 mL min of He makeup. The calibration curve was constructed using a five point
-1
linear fit evenly distributed from 0.2-1.0 mg mL . All samples were diluted 1000 fold in dichloromethane
according to the same dilution scheme.

PBI resin testing


Polybenzimidazole (PBI) was obtained from PBI performance products Inc. The loading capacity of 3-HP
on the resin was determined through measuring the loading curve shown in Figure S33. For these
experiments, 0.8 g (~8 mL wetted volume) of dry PBI resin was stirred in 200-proof ethanol for 20
minutes. The ethanol was vacuum filtered off and the resin slurried into a 25 mL burette in UHP water with

S12
a quartz wool plug at the base of the column. The column was allowed to settle and then rinsed with 10
bed volumes (BV) of UHP water to remove any residual ethanol. 5-6 BV of cation-exchanged and
activated carbon (3 w/v%) treated 3-HP broth was added on top of the resin and drained through at a rate
-1 -1
of 3 BV hr (0.4 ml min ), a rate suggested by Kawabata (68). Effluent samples were taken at intervals
between 0.3-0.6 BV and analyzed for acid concentrations as described in the analytical methods above.
The breakthrough point was determined as the point when the effluent concentration reached 0.1 wt.% of
the target acid. Using the breakthrough point the loading capacity of the resin was determined, in mg of
-1
the target acid g dry resin using Equation (S3) below.

! !""#$!%& ×!!"#$
𝑙𝑜𝑎𝑑𝑖𝑛𝑔 𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦 = (S3)
!!"# !"#$%

In Equation (S3), Veffluent is the total volume in ml of effluent that was collect up to the breakthrough point,
-1
Cacid is the concentration of 3-HP in the broth in mg mL , and mpbi resin is the dry mass of the PBI resin
present in the column.
Elution profiles in Figure S33 were constructed to determine the minimum volume of ethanol required to
completely remove 3-HP from the column. Here 0.8 g of PBI resin was columned in a 25 mL burette as
described above. The resin was loaded with acid by passing cation-exchanged and activated carbon (3
w/v%) treated 3-HP broth through the column to load the resin with acid at the capacity determined from
-1
equation (3) of 230 mg 3-HP g dry PBI resin. 4-5 BV of ethanol elution solvent were passed through the
-1
column at a rate of 3 BV hr . Effluent samples were taken at intervals between 0.3-0.6 BV and analyzed
for 3-HP concentrations as described in the analytical methods above. The minimum volume of eluent
necessary to elute the target acid was determined by the first elution fraction to yield a concentration of 3-
HP below 0.2 wt%.

Process design details


The process design includes feedstock handling and storage, deconstruction of the biomass via
pretreatment and enzymatic hydrolysis, solid/liquid separation of the hydrolysate and its concentration,
biological upgrading of the hydrolysate to 3-HP, recovery of the 3-HP and catalytic upgrading to bio-ACN,
onsite wastewater treatment, and all other required utilities.
Milled biomass enters the process facility as “throat of the reactor ready” where no additional pre-
processing is required prior to the first unit operation. The biomass (with composition outlined in Table S7)
is conveyed to the pretreatment area where it is first treated in a deacetylation step that uses a sodium
-1
hydroxide soaking process (loaded at 14 mg of NaOH g of dry biomass) at 90°C for 1 hour in a set of
parallel batch reactors. This deacetylation step helps to soften the biomass and begins deconstruction by
removing impurities and inerts from the feedstock (such as ash, extractives, and lignin) (64, 69). The
details of the key process parameters and yields for this step are outlined in Table S8.
The remaining solids are conveyed to a series of 8 commercial-scale disc refiners that require a total
-1
power of 200 KWh dry metric ton of biomass. This mechanical refining process further reduces the
particle size of the biomass, as well as deconstructs the biomass to improve the enzymatic conversion of
the biomass to sugars. The whole slurry of refined biomass undergoes enzymatic hydrolysis by a
-1
cellulase enzyme prepared on-site at a loading of 10 mg g glucan. The slurry is fed at a 20 wt% total
solids to a high-solids continuous reactor. The partially hydrolyzed slurry is then batched to one of five
parallel reactors. This process converts glucan, xylan, and arabinan in the solids to monomeric glucose,
xylose, and arabinose, respectively. Table S8 summarizes the key process conditions and yields utilized
in the design for the enzymatic hydrolysis step.
The mixed slurry hydrolysate that is produced is subsequently purified via flocculant-aided solid/liquid
separations to remove any solids in the stream, which would inhibit the aerobic fermentation as well as
limit the ability to concentrate the hydrolysate stream. The clarified stream is then concentrated via a
mechanical vapor recompression (MVR) unit with additional details provided in Biddy et al. and Davis et
al. (22, 70).

S13
The production of 3-HP occurs via aerobic biological conversion of the purified hydrolysate. Yields of the
overall conversion adopted in this process design are provided in Table S8. The maximum theoretical
-1
yield considered for this process design are 2 mol 3-HP mol hexose sugar (glucose) (71, 72), and 1.67
-1
mol 3-HP mol pentose sugar (xylose, and arabinose) (73). This process assumes that the organism is
robust enough to tolerate the low pH of the fermentation broth and that there is no need to add sodium
hydroxide or any buffer to the fermentation broth. Borodina and co-workers have recently demonstrated
biological cultivation of 3-HP at low pH using an engineered strain of Saccharomyces cerevisiae (21).
Thus microorganisms capable of low pH production of 3-HP are already available.
The 3-HP recovery from the broth is achieved in the process outlined in the patent literature for the
recovery of organic acids from fermentation broth (Figure S34). In this design, the low pH broth is passed
through a simulated moving bed (SMB) packed with a PBI resin, which exhibits a high affinity for the acid.
The acid is eluted off of the resin using ethanol and this process allows for 98-100% recovery of the acid
(24). We note that our batch scale chromatography results shown in Figure S33 confirm this 98%
recovery target is met using this resin. We also note that dewatering via a simulated moving bed (SMB)
has significant industrial precedent as the preferred method for dewatering acids from fermentation broth.
For example, lactic acid (a structural isomer of 3-HP) is concentrated from fermentation broth at the
industrial scale into an 88 wt.% solution that is sold on the market (74). Thus, the dewatering process is
modeled to be 98% efficient with a resulting product stream of 5 wt% organic acid, 2 wt% water, and 93
wt% ethanol based on the patent details (24) and our experimental data shown in Figure S33. The
recovered stream is further concentrated (to a 50/50 wt% mixture of ethanol and 3-HP) via distillation with
the recovered ethanol being recycled to the SMB process (Figure S34).
The product mixture is then sent into a CSTR reactor and a reactive distillation process where, at a mild
temperature of 90°C and in the presence of a sulfuric acid catalyst, the mixture is esterified to produce an
ethyl acrylate product (Figure S35). We note that this is the same industrial process previously developed
in the 1950’s for the production of acrylate esters from β-propiolactone (BPL) using reactive distillation
(25). The chemistry of this process is identical to that using 3-HP. In the previously developed industrial
process, BPL is added to an ethanol solution acidified with sulfuric acid in a CSTR. BPL rapidly ring opens
to form 3-HP and then esterifies and dehydrates to form an acrylate ester (25). The parameters described
in the patent literature, and used in the model, for the CSTR operation are a 3 hour residence time, at
90°C, with 2-4 wt.% H2SO4 used as the catalyst. Recovery of the resulting product and overall energy
demands from the reactive distillation process is adopted from a recent publication (26). Overall, this
process was reported to achieve a 95% yield of ethyl acrylate from BPL in the 1950s (25). Given that
chemical process technology has significantly advanced since then, the 95% yield from this process is
considered significantly de-risked. The resulting ethyl acrylate is recovered via distillation at a product
purity of 98 wt% and is further upgraded via the nitrilation process.
The ethyl acrylate is co-fed with ammonia (at a 8:1 molar ratio of ammonia to ethyl acrylate) in a fluidized
bed reactor system at 310°C and 1 atm. Within, the ethyl acrylate is converted to acrylonitrile (Figure 3B
of the main text) utilizing a titanium oxide catalyst at a 10:1 catalyst to ethyl acrylate loading. Overall
conversion of the ethyl acrylate is 98% to ACN while 2% is lost to coke. The coke is combusted at
elevated temperatures (550°C), which helps to heat the catalyst and provide heat for the endothermic
reaction. We assume a 5% heat loss in the overall integrated system. A 2% loss of catalyst is also
assumed, consistent with previous designs (75).

The reactor outlet is quenched (utilizing a chilled stream of dimethyl sulfoxide or DMSO), allowing a near
complete recovery of the ammonia as a vapor, which is recycled to the front of the acrylonitrile production
reactor (Figure S42). Any entrained ammonia is neutralized with sulfuric acid via an inline mixer; the
resulting ammonium sulfate salt is recovered and sent to wastewater treatment (Figure S43 and Figure
S44).

The remaining mixture of ethanol, acrylonitrile, and water forms a ternary azeotrope (Figure S42); the
recovery that meets a 97 wt% purity of acrylonitrile is achieved through a distillation/decanting system
(Figure S43). The overhead of the distillation column is rich in ACN while the bottom of the column is rich

S14
in ethanol (as a mixture with water) which is further distilled to the azeotrope and then separated via
molecular sieve (Figure S44) (76). The recovered ethanol is cooled and then recycled to the 3-HP
recovery section while the water is sent to wastewater treatment. The acrylonitrile product is recovered at
a purity of 97% from the decanting system. (Figure S43). All key process parameters are reviewed for the
upgrading strategy in Table S8.
Supporting processes, including wastewater treatment, utilities (including cooling water, chilled water
plant and instrument air, and process water), and the combined heat and power systems (consisting of a
combustor, boiler, and turbogenerator) are also incorporated in the design and economics with detailed
descriptions found in Davis et al. (70).

Economic Analysis Approach


Aspen Plus process models were developed based on the conceptual process designs described above
(Figure S34 through Figure S44) and key process parameters outlined in Table S8. The material and
energy flows in the process and the overall variable and fixed operating costs, as well as the capital costs,
were then estimated for the integrated design and the calculated flow rates. The capital costs, which are
outlined in Table S9, are based on vendor designs when available. The details of these equipment
designs have been published by Biddy et al., Humbird et al., and Davis et al. (22, 70, 76, 77).
To estimate the minimum selling price of the bio-ACN, a discounted cash flow analysis was applied under
a specific set of financial assumptions. These financial assumptions are reviewed in Table S10, which are
based on a mature nth-plant and consistent with prior published work (22, 70, 76-81).

Economic Analysis Results


Utilizing the design and economic basis described above, a minimum selling price (MSP) of the bio-ACN
was estimated to be $0.89/lb. This estimate is based on a 2000 dry metric ton/day facility producing 339.2
MM lb of bio-ACN per year. The cost breakdown for the major process areas is provided in Figure S45
and details provided in Table S9, Table S11, and Table S12. There are two process areas that contribute
over 50% to the total MSP. First, biomass feedstock prices at $84.45/dry ton (as outlined in Table S12)
contribute $0.18/lb to the bio-ACN MSP. Second, conversion of the 3-HP to bio-ACN contributes $0.33/lb
to the bio-ACN MSP. Ammonia consumption for the bio-ACN production impacts the MSP by $0.09/lb.
Capital costs of the ACN production contributes to roughly $0.15/lb of the MSP. Overall, 42% of this
capital cost is attributed specifically to the recovery of 3-HP and an additional 30% of the capital cost is
the result of the various reactors in the 3-HP catalytic upgrading process.

Sensitivity Analysis
To further evaluate the impact that design and cost assumptions have on the economics of the integrated
design, a series of sensitivity analyses were performed. As shown in Figure S46, two of the key drivers for
economics of the process are the amount of sugar produced and converted to 3-HP. If xylose conversion
in enzymatic hydrolysis is eliminated, the price of the ACN increases by nearly $0.36/lb, while if glucose
conversion is reduced from 80% to 70%, the ACN price increase by $0.08/lb. Eliminating arabinose
production (from the baseline 80%) has less of an impact at an increase in cost of $0.05/lb ACN. If sugar
production could be increased for all cases, this would lower the ACN cost by roughly $0.10/lb for the
case of glucose and $0.05/lb for xylose. Arabinose impact is negligible given its low concentration. This
study considered the variation of enzyme loading in enzymatic hydrolysis. If enzyme loading were
doubled to 20 mg enzyme/g glucan, the ACN selling price increases to $0.93/lb while reduction of enzyme
loading to 5 mg would lower the cost to $0.87/lb.
Additional drivers to the process economics are centered on the upgrading of 3-HP in the production of
ACN. In this series of sensitivity analyses this study considered the impact around conversion yields,
recovery efficiencies and cost associated with the required raw material feedstocks for the process. When
looking at overall conversion yields, reduction of the yield in the acrylonitrile upgrading process from 98%
to 90% increases the ACN price to $0.95/lb. If the 3-HP conversion to ethyl acrylate is reduced from 100%
to 90%, the estimated ACN price increases to $0.98/lb due to the lower overall ACN yields. Another factor
that affects the process yield is the efficiency of recovery and this was explored by considering the 3-HP

S15
recovery efficiency. If the 3-HP is only recovered at 90% (instead of the 98% modeled from the patent
literature (24) and our results in Figure S33), the ACN price increase by $0.07/lb. When considering the
overall impact of cost on this upgrading process, if the ammonia price is doubled from the current price
assumed the ACN price increases by $0.09/lb, while halving the price of ammonia to $250/ton would
reduce the ACN price by $0.05/lb. On the other hand, due to the rigorous approach for recovery of the
ethanol with >99.995% of the ethanol being recovered, the price of ethanol has limited effect (<$0.01/lb)
on the ACN minimum selling price. The cost associated with the ACN upgrading reactor is roughly
$33MM installed for the baseline model. If the price is doubled, the overall price of ACN increase by
$0.03/lb while cutting this cost in half only influences the ACN price by $0.01/lb.
Financial economic factors were also explored by looking at variation of internal rate of return (IRR) of the
process. The current baseline assume a 10% IRR, however, this assumption may be too low for
commodity chemical production as these processes tend to garner a higher IRR. If a 15% IRR is targeted
for the production of ACN, the chemical price increases by $0.13/lb. On the other hand, we also explore
the breakeven price of the process (without considering profits) by setting the IRR to 0% which resulted in
an ACN price of $0.65/lb (Figure S46).
This analysis also explored the use of a first generation sugar feedstock. For comparison to the
lignocellulosic process, Number 11 commodity (raw) sugar, which is primarily sucrose, was used. The
global Number 11 commodity sugar has a current future prices cited at $0.15/lb and a 10 year average
price of around $0.18/lb (82). This process assumes the yeast that produce the 3-HP can also convert
sucrose (a disaccharide), such that additional enzymes or processing of the sugar is not needed for
production. The ability for many yeast strains to consume sucrose is well documented (83). All other
downstream conversion of 3-HP to ACN is kept constant with the overall yield of ACN set to be the same
as the lignocellulosic process. Additionally, the overall process included all required supporting equipment
for the integrated design such as wastewater treatment, steam generation (which is supplied by natural
gas), and all cooling and chilled water needs. Overall, for this first generation sugar case the capital cost
of the process decrease by nearly 30% when compared to the lignocellulosic case with a total capital
investment of $620MM. As noted, natural gas is required to supply the heat needed for the process which
adds roughly $0.06/lb to the ACN price for the commodity sugar case considered. Despite this additional
operating cost for natural gas costs, the combined with the reduction in feedstock costs and the capital
costs results in an ACN price ranging between $0.72/lb (at $0.15/lb sugar) to $0.76/lb (for a $0.18/lb
sugar). In the baseline design for the lignocellulosic feedstock, the estimated minimum sugar selling price
for the concentrated second generation (2G) sugar is $0.19/lb.
Finally, we note that a $0.05/lb reduction in the selling price for ACN can be achieved if catalyst
development for the nitrilation reaction progresses such that a yield of 98% AN can be maintained for a
time on stream of 30 minutes at the full scale reaction conditions (310°C, 90.0 kPa NH3, and 11.26kPa
EA). This would enable the reaction to be run in multiple packed bed reactors that cycle through
regeneration and reaction sequences. The $0.05/lb price reduction is due to the lower energy and capital
requirement costs for operating fixed bed systems compared to a fluidized system. There would also
much less catalyst attrition in a packed bed system. Given that ACN is a commodity chemical and
achieving the lowest price possible is key for industrial adoption, we feel there are substantial research
opportunities in developing more active and fouling-resistant catalysts for this reaction.

Life cycle assessment


LCA is a tool for a wide range of sustainability indicators, such as ozone depletion and terrestrial toxicity.
This study focuses only on global warming potential (GWP, represented in grams of CO2-equivalents
(CO2e)) using a 100-year GWP. The goal of this LCA study is to estimate the life cycle greenhouse gas
(GHG) emissions associated with the production of biomass-derived ACN. The functional unit is 1 kg of
bio-ACN produced (i.e., field-to-gate [FTG]). For this analysis, we accounted for all the stages in the life
cycle of the bio-ACN, including the biomass feedstock production and logistics, and bio-ACN production.
Figure S47 shows the FTG life cycle stages for the current biomass-derived ACN bioproduct pathway.
Note that the bioproduct use phase and the end-of-life phase were not included in the system boundary in
this study as bio-ACN will be further processed to renewable fiber which is the final product.

S16
SimaPro v.8.0.2 software was used to develop and link units quantifying life cycle impacts (84). GHG
basis values for corn stover production and logistics, jet fuel transportation and end use, electricity and
hydrogen were applied consistent with the basis from Argonne National Laboratory’s GREET model
software (85). Ecoinvent v.2.2 database and the U.S. Life Cycle Inventory (LCI) processes were used to
fill the data gaps (86). The Ecoinvent processes are modified to reflect U.S. conditions and the U.S. LCI
processes are adapted to account for embodied emissions and fossil energy usage (87). The material
and energy flows or life cycle key inventory data of the conversion step capture the impacts of input raw
materials, and outputs, such as emissions, wastes, and co-products as predicted by the process models
(Table S13).
The overall GHG emissions from the biomass-derived ACN process were determined to be 2.87 kg
CO2e/kg, corresponding to 14.1% improvement compared to fossil-based ACN. The life cycle GHG
emission for the fossil-derived ACN is 3.34 kg CO2e/kg and the GHG reduction was calculated using
equation (S4):

!"!!"##$%!!"# !!"!!"#!!"#
𝐺𝐻𝐺!"#$%&'() (%) = ∗ 100 (S4)
!"! !"##$%!!"#

For comparison, LCA for ACN derived from first generation (1G) sugars was also performed. There are
variety types of 1G-sugars, including glucose from corn starch, sucrose from sugar cane, and sugar beet
molasses (predominantly sucrose but also glucose and fructose). For the 1G LCA, sucrose from sugar
cane was chosen as the feedstock for the 1G renewable acrylonitrile production. We accounted for all the
stages in the life cycle of the 1G bio-ACN, including the sugar production and logistics (88), and bio-ACN
production, as shown in Figure S48. The life cycle key inventory data of the 1G conversion are
summarized in Table S14. The overall GHG emissions from the biomass-derived ACN process were
determined to be 3.02 kg CO2e/kg (Table S15).

SUPPLEMENTARY TEXT
Further discussion of computational results
Gas-phase thermochemistry
Nitrilation enthalpies of formation and enthalpies of reaction were estimated using both experimentally-
available values and computational methods. To the best of our knowledge, there is no experimentally-
found enthalpy of formation of ethyl 3-HP, so it was approximated using a value for ethyl lactate (EL), -
-1 -
695.1 kJ mol (89). However, the error for this value is cited to be less than 25% (or less than 173 kJ mol
1 -1
). Furthermore, the NIST-cited ethyl acrylate (EA) values range between -327.8 and -354.2 kJ mol .
Alternatively, computational methods were used to estimate gas-phase thermochemistry. The results are
presented in Table S2. As expected, the largest discrepancies between experimental and computational
values arise from the imprecision of experimental EA and ethyl 3-HP values. Ethyl 3-HP dehydration is
-1
found to be only slightly endothermic using computational methods (+36 to +39 kJ mol ) compared with
-1
experimental estimates (+99 to +126 kJ mol ). The overall ethyl 3-HP nitrilation is found to be significantly
-1 -1
endothermic using either computational (+127 to +154 kJ mol ) or experimental (+196 to +203 kJ mol )
estimates.
Ethyl 3-HP dehydration mechanism
We investigated three possible pathways for ethyl 3-HP dehydration: (1) a dissociative pathway, (2) an E2
pathway with a single-step dehydroxylation and deprotonation to form ethyl acrylate, and (3) an E1
pathway, forming a carbocation intermediate prior to deprotonation. A carbocation intermediate was not
found to be stable on TiO2(101). The other two pathways are presented in Figure S5. All reaction
energetics associated with the dissociative and E2 pathways are presented in. Key intermediates are
shown in Figure S6. The minimum-energy path (Figure S7) was determined to follow the E2 mechanism,
-1
with the highest barrier estimated for the C-OH and C-H scission at 112 kJ mol . In contrast, the C-O and
-1
C-H scission along the dissociative path was found to be 46 kJ mol more endothermic. Water formation
-1
was found to be facile with a barrier of 26 kJ mol .

S17
Aminolysis mechanism
The energetics associated with three mechanisms for ethyl acrylate aminolysis are presented in Figure
S8: (a) concerted single-step mechanism, (b) stepwise mechanism and (c) a mechanism involving
surface acrylate formation. The corresponding reaction energetics are presented in Table S4. The key
stable intermediates along aminolysis paths are shown in Figure S9. Gibbs free energy profiles (310°C)
for concerted, stepwise, and acrylate paths are presented in Figures S10, S11, and S12, respectively.
Previous gas-phase ab initio investigations suggest that an amide can form via either a concerted
reaction involving simultaneous NH3 insertion and C-O scission or via a stepwise mechanism involving
NH3 insertion to form a neutral intermediate followed by hydrogen transfer and C-O scission to form an
alcohol (90, 91). Our DFT calculations suggest a stepwise ethyl acrylate aminolysis with an added partial
NH3 dissociation step for reaction (2) in Scheme 1 of the main text. Along this path, H-NH2 scission has
-1
the highest barrier of 111 kJ mol . Subsequent mechanistic steps are exothermic and kinetically facile:
NH2 insertion to form a stable intermediate (1-amino-1-ethoxy-2-propen-1-ol) and C-O scission to form
-1
acrylamide and ethanol have respective barriers of 36 and 21 kJ mol (Figure S11). These results are in
line with literature showing that partial NH3 decomposition occurs over dehydroxylated TiO2 (92). An
alternative stepwise mechanism for reaction (2), involving C-O scission first to form surface acrylate and
ethoxy species with subsequent NH2 insertion, was found to be potentially relevant; however, further
microkinetic analysis is required (and will be performed in future work) to distinguish the stepwise
mechanisms, both of which hinge on partial NH3 dissociation (Figure S12).
Amide dehydration
We computed the energetics for four potential acrylamide dehydration paths: (A) surface-mediated
tautomerization to acrylamide enol form followed by either a sequential N-H and C-OH scission or (B)
simultaneous C-OH and N-H scission, (C) ethanol-forming single-step tautomerization followed by C-OH
and N-H scission, and (D) dissociation of acrylamide without tautomerization. These pathways are shown
in Figure S13. The corresponding energetics are summarized in Table S5. The key stable intermediates
along amide dehydration paths are shown in Figure S14. Gibbs free energy profiles (310°C) for paths A,
B and C are presented in Figure S15, Figure S16, and Figure S17, respectively. The stepwise
dissociation of acrylamide (path D) was not found to be energetically favorable, since dehydrogenation
-1 -1
and deoxygenation steps are highly endothermic (86 kJ mol and 156 kJ mol , respectively).
We find that acrylamide is likely to dehydrate by first undergoing tautomerization to its enol form. One
potential route involves surface-mediated tautomerization, in which N-H scission is followed by O-H
-1 -1
formation. This path is endothermic (46 kJ mol ), but has low barriers of 40 and 23 kJ mol , respectively.
Alternatively, ethanol formation during aminolysis can facilitate acrylamide tautomerization via hydrogen
transfer. In this step (AD10), acrylamidic oxygen picks up surface hydrogen while hydrogen from the
-1
amine is picked up by the ethoxy (EtO*) species. This significantly lowers the reaction energy to 6 kJ mol
-1
and makes the reaction barrier negligible (7 kJ mol at 310°C in Figure S17). An enol form of the amide
along its dehydration mechanism has been proposed both experimentally (93) as well as computationally
(94, 95) and proves to be the likely pathway for amide dehydration in this work as well.
-1
The dehydration of the enol proceeds through N-H scission first (AD6) with a barrier of 92 kJ mol (101 kJ
-1 -1 -1
mol at 310°C), followed by facile C-OH scission (AD11) with a barrier of 32 kJ mol (38 kJ mol at
-1
310°C). The simultaneous N-H and C-OH scission is not likely, showing a higher barrier of 125 kJ mol
-1
(134 kJ mol at 310°C).

Further discussion of scaled up process model used in the economic analysis


In the scaled version of the process (Figure S34 to Figure S44), the catalytic nitrilation chemistry is
modeled in pure ammonia vapor consistent with a previous report of pilot scale acetonitrile production
from acetic acid (Table S8) (96). In the data shown in the main text N2 dilution was used only because the
online FTIR system can not quantify with analyte concentrations >15% vol/vol. Thus, to avoid detector
saturation N2 was used as a carrier gas diluent. However, catalyst deactivation studies (Figure S38 and
Figure S39) were fit empirically following Equation S5 as a rate expression for catalyst deactivation.

S18
! ! !
!
= exp −𝑎𝑃!! 𝑃 ! 𝑡 !!!!! !!"
! !"
(S5)
!!

The best fit values of a, b, c, d, e, f are found as:

𝑎 0.2264
𝑏 -1.3559
𝑐 1.8758
𝑑 0.5041
𝑒 0.2216
𝑓 -0.1740

Using the reactant partial pressure from the design parameters for the full-scale economic model PNH3 =
90.0 kPa and the PEA = 11.26 kPa (Table S8) in Equation (S5) finds that a 2% loss in yield occurs at 0.5
minutes (Figure S41). Assuming the catalyst and reactants are well mixed in a fluidized bed, every 0.5
minutes the entire bed of catalyst must be circulated in the riser reactor to ensure > 98% yield of
ACN.

In the first run in Figure 4, the entire 43 g of catalyst has lost 8% of its activity (A/A0) and the TG-FTIR
results during regeneration of this entire catalytic bed find that 1.14 g of mass is burned off of the entire
TiO2 catalytic bed at 550°C. Catalyst deactivation studies were repeated but to varying degrees of A/A0
and the amount of coke burned off during regeneration was quantified via TGA-FTIR. Results of g coke
burned off / g catalyst versus A/A0 from these experiments is plotted in Figure S40. A best fit line to this
data set was found to be y = -0.06x + 0.062. From this linear fit to the results in Figure S40, we estimate
-1
that at 98% activity ~ 0.003 g of coke g TiO2 must be burned off in the regenerator.

We also note that handling of acrylates at elevated temperatures can present an explosive polymerization
hazard. However, acrylate vapors are stable (97). The dangers of violent polymerization events are
specific to collection of the uninhibited acrylate when it is cooled and condensed and this can be achieved
safely by condensing the acrylate vapors simultaneously with an inhibitor (98). In the process described in
this work, the volatilization of ethyl acrylate into the reactor does not present a polymerization hazard
since the vapors are stable. However, care is taken for the safe condensation and collection of produced
acrylonitrile. This was demonstrated and achieved in our lab by bubbling the hot product vapors through
chilled DMF (dimethylformamide) at 0°C that contains a polymerization inhibitor (4 hydroxy TEMPO at
100 ppm). This stripping system safely condenses and collects acrylonitrile, water, and ethanol into DMF,
which can then be distilled to collect neat inhibited products. Additionally, the aprotic nature of DMF allows
the ammonia gas to bubble through where it is then collected in a gas accumulator for recycling to the
reactor. The scaled process is identical except that DMSO is used in place of DMF. We note that the
decantor-distillation setup in Figure S43 also contains 100 ppm of 4 hydroxy TEMPO as an inhibitor to
stabilize the final acrylonitrile product.

S19
SUPPLEMENTARY FIGURES
N2
Flowrates!
TiO2 N2: 2000 sccm (8.92E-4 mol/min)!
NH3

H2O NH3: 120 sccm (5.35E-3 mol/min)!


Ethyl 3-HP: 0.077 ml/min (6.87E-4 mol/min)!

!
5 g of TiO2 catalyst!
!
Dehydration +
Nitrilation
mol / min

WHSV = 0.97 hr-1! TiO2


Contact time ~ 0.04 sec!
NH3 to EA molar ratio = 8:1!
!

EtOH

H 2O

Temperature (°C)
Figure S1. Complete reaction conditions and dataset (with molar units) for observed reaction products for
the data presented in Figure 1A of the main text.

S20
H H 2O
O H


DEHYDRATION
H OH



Figure S2. Reaction mechanisms predicted from DFT calculations for the dehydration of ethyl 3-HP to
form gaseous ethyl acrylate and water (Reaction (1) in Figure 1B).

S21
Figure S3. Profile (left) and top (right) view of anatase TiO2(101) model used in DFT modeling. The layer
relaxed in all optimizations (atoms closest to the surface) is shown as spheres; the layers kept in their
optimized bulk positions are represented with a wireframe.

S22
Figure S4. Convergence tests performed in this work: (A) TiO2(101) slab energy as a function of k-point
mesh, (B) NH3 adsorption energy difference as a function of cutoff energy, and (C) NH3 adsorption energy
difference with the number of TiO2(101) bulk layers. The inset images, from left to right, represent one,
three, and six bulk TiO2(101) layers, respectively.

S23
Figure S5. Ethyl-3-hydroxypropanoate (E3HP) dehydration mechanisms over TiO2: (A) dissociative
pathway and (B) concerted E2 pathway. Purple arrows represent the electron flow for the reactions
following the intermediate. The reactions associated with the dissociative pathway are E3HP adsorption
(D1), O-H scission (D4), C-O and C-H scission to produce ethyl acrylate (D7) and a series of H2O
formation steps not shown here. The E2 pathway proceeds through the same adsorption, followed by
simultaneous C-OH and C-H scission to form ethyl acrylate (D8), followed by H2O formation (D6).

S24
Figure S6. Key intermediates in ethyl-3-HP dehydration over TiO2(101). Surface atoms are shown as a
wireframe; surface Ti and O (Os) atoms involved in bonding are shown as spheres for clarity. Bond
distances are presented in Å.

S25
Figure S7. Gibbs free energy diagram for ethyl-3-HP dehydration over TiO2(101) via an E2 mechanism at
310°C. Free energy levels (kJ/mol) are shown in blue. Transition states are shown in the inset images:
simultaneous C-OH scission and deprotonation to form ethyl acrylate (left) and water formation (right).
Bond lengths (Å) are shown in black.

S26
Figure S8. Ethyl acrylate (EA) aminolysis mechanisms considered in this work: (A) concerted, single-step
pathway, (B) stepwise pathway, and (C) surface acrylate pathway. The concerted pathway involves NH3
and EA coadsorption, followed by a single-step ethanol and acrylamide formation (A5). The stepwise
pathway involves EA conformational change (A12), partial NH3 decomposition (A8), and subsequent NH2
insertion (A6) to form a stable 1-amino-1-ethoxy-2-propen-1-ol surface intermediate (AEP*). This
intermediate undergoes rapid C-O scission (A9) to form acrylamide, followed by ethanol formation (A11).
The third pathway involves direct C-O scission (A10) to form a surface acrylate intermediate, partial NH3
decomposition (A8), and subsequent NH2 insertion (A7).

S27
Figure S9. Key intermediates in ethyl acrylate aminolysis over TiO2(101). Surface atoms are shown as a
wireframe; surface Ti and O (Os) atoms involved in bonding are shown as grey and dark red spheres,
respectively, for clarity. Bond distances are presented in Å.

S28
Figure S10. Gibbs free energy diagram for ethyl acrylate aminolysis over TiO2(101) via the concerted
mechanism at 310°C. Free energy levels (kJ/mol) are shown in blue. Transition state for the simultaneous
H-NH2 scission, C-O scission, NH2 insertion and O-H formation is shown in the inset images. Bond
lengths (Å) are included in black.

S29
Figure S11. Gibbs free energy diagram for ethyl acrylate aminolysis over TiO2(101) via the stepwise
mechanism at 310°C. Free energy levels (kJ/mol), which are referenced to clean TiO2(101) and gas-
phase ethyl acrylate (EA) and NH3, are shown in blue. The first two barrier-less steps represent the
sequential adsorption of NH3 and EA (A2 and A1’), respectively. Transition states are shown in the inset
images, from left to right: ethyl acrylate conformational change (A12), H-NH2 scission (A8), NH2 insertion
(A6), C-O and O-H scission (A9), and O-H formation (A11); bond lengths (Å) are included in black. The
last two barrier-less steps are EtOH and acrylamide desorption (A4’ and A3), respectively.

S30
Figure S12. Gibbs free energy diagram for ethyl acrylate aminolysis over TiO2(101) via the acrylate
mechanism at 310°C. Free energy levels (kJ/mol), which are referenced to clean TiO2(101) and gas-
phase ethyl acrylate (EA) and NH3, are shown in blue. The first two barrier-less steps represent the
sequential adsorption of NH3 and EA (A2 and A1’’), respectively. The next step (-58 to 35 kJ/mol) is a
conformational change/diffusion of EA (A13). Transition states are shown in the inset images for the next
four reactions, from left to right: C-O scission (A10), H-NH2 scission (A8), O-H formation (A11), and NH2
insertion (A7); bond lengths (Å) are included in black. The last two barrier-less steps are EtOH and
acrylamide desorption (A4’ and A3), respectively.

S31
Figure S13. Acrylamide dehydration mechanisms considered in this work: surface-mediated
tautomerization to acrylamide enol form followed by (A) sequential N-H and C-OH scission or (B)
simultaneous N-H and C-OH scission, (C) ethanol-forming single-step tautomerization followed by N-H
and C-OH scission and (D) dissociation of acrylamide without tautomerization via a series of N-H and C-O
scission steps. Surface-mediated tautomerization for paths A and B involve N-H scission (AD5) followed
by O-H formation (AD13). Path A further involves N-H scission (AD6) and C-OH scission (AD11). Path B
follows a simultaneous N-H and C-OH scission (AD8). Path C involves hydrogen transfer to
simultaneously form ethanol and tautomerize acrylamide to its enol form (AD10). The co-adsorbed
surface ethanol is not shown along this path. Path D involves direct C-O scission (AD9).

S32
Figure S14. Key intermediates in acrylamide dehydration over TiO2(101). Surface atoms are shown as a
wireframe; surface Ti and O (Os) atoms involved in bonding are shown as grey and dark red spheres,
respectively, for clarity. Bond distances are presented in Å.

S33
Figure S15. Gibbs free energy diagram for acrylamide dehydration over TiO2(101) via surface-mediated
tautomerization and stepwise N-H and C-OH scission at 310°C. Free energy levels (kJ/mol), which are
referenced to clean TiO2(101) and gas-phase acrylamide (AAm), are shown in blue. The first barrier-less
steps represent AAm adsorption (AD1) and conformational change (AD4), respectively. Transition states
are shown in the inset images, from left to right: N-H scission (AD5), O-H formation (AD13), second N-H
scission (AD6), C-OH scission (AD11) and H2O formation (AD15); bond lengths (Å) are included in black.
The last barrier-less step is H2O desorption (AD3).

S34
Figure S16. Gibbs free energy diagram for acrylamide dehydration over TiO2(101) via surface-mediated
tautomerization and a single-step N-H and C-OH scission at 310°C. Free energy levels (kJ/mol), which
are referenced to clean TiO2(101) and gas-phase acrylamide (AAm), are shown in blue. The first barrier-
less steps represent AAm adsorption (AD1) and conformational change (AD4), respectively. Transition
states are shown in the inset images, from left to right: N-H scission (AD5), O-H formation (AD13), a
concerted N-H and C-OH scission (AD8) and H2O formation (AD15); bond lengths (Å) are included in
black. The last barrier-less step is H2O desorption (AD3).

S35
Figure S17. Gibbs free energy diagram for acrylamide dehydration over TiO2(101) via direct H-transfer
tautomerization followed by step-wise N-H and C-OH scission at 310°C. Free energy levels (kJ/mol),
which are referenced to acrylamide (AAm) coadsorbed with ehtoxy (EtO) and H species TiO2(101), are
shown in blue. Transition states are shown in the inset images. The first involves AAm tautomerization to
its enol form while simultaneous forming ethanol (AD10). The next transition states, from left to right, are:
N-H scission (AD6), C-OH scission (AD11) and H2O formation (AD15). Bond lengths (Å) are included in
black. The last barrier-less step is H2O desorption (AD3).

S36
!! TiO2
(1) + H2O

(1) 100 ± 4 kJ mol-1 Flowrates into reactor!


N2: 2000 sccm (8.92E-2 mol/min)!
Ethyl 3-HP: 0.077 ml/min (6.87E-4 mol/min)!
!
!
TOF ( h-1)

Reactor parameters!
Temperature ranged from: 140-160°C!
2.00 g TiO2!
!

1/T (K-1)
Figure S18. Arrhenius plot for the dehydration of the 3 position –OH group in ethyl 3-HP to form ethyl
acrylate and water. Substrate conversion is <15%.

S37
TiO2
+" NH3 +" +""H2O
103 ± 12 kJ mol-1
Flowrates into reactor!
TOF ( h-1)

N2: 1900 sccm (8.47E-2 mol/min)!


Ethyl acrylate: 0.077 ml/min (7.23E-4 mol/min)!
NH3: 120 sccm (5.35E-3 mol/min)!
!
Reactor parameters!
"
Temperature ranged from: 290 – 330°C!
2.00 g TiO2!
!

1/T (K-1)
Figure S19. Arrhenius plot for the nitrilation of ethyl acrylate to form ethanol, water and acrylonitrile. This
-1
analysis yields an apparent activation energy for the nitrilation of ethyl acrylate of 103 ± 12 kJ mol .
Substrate conversion is <15%.

S38
N2
EtOH

Flowrates into reactor 1!


N2: 1923 sccm (8.58E-2 mol/min)!
EtOH Ethyl 3-HP: 0.077 ml/min (6.87E-4 mol/min)! Dehydration
Ethanol: 0.08 ml/min (1.37E-3 mol/min)! TiO2 260°C
!
Reactor 1 parameters!
WHSV: 0.5 hr-1!
mol / min

Temp: 260 °C! NH3


H 2O Contact time ~0.08 sec!
12g TiO2!
!
Flowrates into reactor 2! Nitrilation
!
N2: 1923 sccm (8.58E-2 mol/min)!
NH3: 200 sccm (8.92E-3 mol/min)!
TiO2 315°C

Ethyl 3-HP: 0 mol/min!


Ethanol: 1.37E-3 mol/min!
Ethyl acrylate: 6.87E-4 mol/min!
H2O: 6.87E-4 mol/min!
!
Reactor 2 parameters!
TiO2 WHSV: 0.11 hr-1!
Temp: 315 °C! H 2O
Contact time ~0.32 sec!
EtOH
43 g TiO2!
TOS (hr)
Figure S20. Complete dataset and conditions used for the results shown in Figure 2 of the main text. The
graph displays all reaction products measure in molar units. Complete reaction conditions used are given
in the grey box.

S39
Flowrates N2
N2: 1923sccm (8.92E-4 mol/min) EtOH
Ethanol 0.08 ml/min (1.37E-3 mol/min)
Ethyl 3-HP: 0.077 ml/min (6.87E-4 mol/min)

TiO2
12 g of TiO2 catalyst
Dehydration
TEMP: 260 °C TiO2
% yield

WHSV = 0.5 hr-1


Contact time ~ 0.08 sec

EtOH

H 2O

TiO2
+ H2O +!

TOS (hr)
Figure S21. High-conversion experiments were performed to maximize the yield of ethyl acrylate formed
when dehydrating the primary alcohol on ethyl 3-HP. Here ethyl 3-HP was passed over the TiO2 catalyst
at 260°C with a 2:1 molar excess of ethanol. The excess ethanol was needed to suppress the equilibrium
formation of acrylic acid, see reaction shown in the bottom right. In the results shown in the graph, 100%
yield of ethyl acrylate was obtained with 100% mass balance using the parameters listed in the grey box.

S40
Ethyl 3HP Nitrilation Reaction! ΔH 0rxn!
TiO2
+" NH3 310 °C
+" +""2H2O +203 kJ/mol!

Figure S22. Reaction enthalpy for the dehydration and nitrilation of ethyl 3-HP to form acrylonitrile water,
and ethanol. The reaction enthalpy was calculated from the values in Table S2. (+179.7+-234+2*-241.8)-
(-45.9+-695.1) = +203 kJ/mol. Note here that the heat of formation of ethyl 3-HP is not available so in its
place we use the heat of formation of ethyl lactate, a structural isomer.

S41
TiO2
Fresh catalyst Spent catalyst Regenerated catalyst
Figure S23. Fresh, spent and regenerated TiO2 catalyst used for the experiments shown in Figure 2.

S42
TiO2

Signal (a.u.)
regenerated

fresh

Wavenumber (cm-1)
Figure S24. DRIFTS spectrum of pyridine on fresh (green) and regenerated (blue) TiO2.

S43
TiO2

Vol. lAds. (cm3 g-1)


Fresh TiO2 – 49 m2 g-1

Spent TiO2 – 27 m2 g-1


Regen TiO2 – 50 m2 g-1

P/P0
Figure S25. Nitrogen physisorption of fresh and spent TiO2 with BET surface areas indicated in the
legend.

S44
TiO2
anatase reflection

rutile reflection

Intensity (a.u.)
fresh

spent

regenerated

°2θ
Figure S26. X-ray diffractogram of fresh, spent, and regenerated TiO2.

S45
TGA-FTIR
Mass loss

FTIR signal
% mass Mass loss
derivative

H2O
signal

CO2
signal

Temperature (°C)
Figure S27. TGA-FTIR results during coke burnoff of the spent catalyst shown in Figure S23. Note here
no NOx was observed. Only CO2 and water was observed.

S46
GLUCOSE

XYLOSE GALACTOSE
glycolysis
ARABINOSE

Phosphoenolpyruvate NADH NAD+


ADP (R)-LACTATE
pykF
pykA ATP ΔldhA

ΔpoxB PYRUVATE tdcE CO2


ΔpflB fdhF
FORMATE
aceEF hycEBG
CO2 lpd hycFCD H2

eutD adhE
ΔackA Δpta ΔadhE adhP
ACETATE Acetyl-phosphate Acetyl-CoA Acetaldehyde ETHANOL
biotin
ATP ADP NADH NAD+
accABCD ATP + HCO3-
CoA
ADP + Pi
fabI
Malonyl-CoA (fatty acid synthesis)

NADPH
mcr
NADP+

Malonate semialdehyde

NADPH
ydfG
NADP+

3-HYDROXYPROPIONIC ACID

Figure S28. Metabolic pathway for 3-HP production in recombinant E. coli BGHP strain.

S47
Concentration (g/L) Glucose

OD

ln OD (600 nm)
glucose

acetate

Time (hr)
Figure S29. Fed-batch DO-stat cultivation of the E. coli BGHP strain using glucose as a carbon source.

S48
3-HPA! Corn stover
glucose acetate! hydrolysate
OD!
Concentration (g/L)

glucose!

ln OD (600 nm)
OD xylose!
arabinose!
acetate galactose!

xylose

arabinose

galactose

Time (hr)
Figure S30. Fed-batch DO-stat cultivation of the E. coli BGHP strain using DDAP-EH corn stover
hydrolysate.

S49
STEP 1 STEP 2 STEP 3 STEP 4 STEP 5
fermentation filtration activated carbon dewatering EtOH dissolution

Color
removal Evaporate to
dryness leaving
salts and sugars
pH = 7!
0.2 μm 10 kDa
filter filter
ethanol

cells + proteins spent


debris carbon
sugars

STEP 6 STEP 7 STEP 8 STEP 9


salt breaking esterification distillation DCVC

Add Purified !
H2SO4 ester!

ethanol ester + alcohol Crude !


solution! ester!

[NH4]2SO4

Figure S31. Cartoon depiction of the nine-step procedure used for separating ethyl 3-HP from
fermentation broth. The process is described in detail in materials and methods section. The recovered
ethyl 3-HP was converted to acrylonitrile as described in the main text and in Figure 2.

S50
Figure S32. Pictures of the glucose fed fermentation broth as it was processed through the nine-step
procedure depicted in Figure S31.

S51
Influent:
A. 20 g/L 3-HPA

PBI resin
Effluent conc. (g/L)

Loading
capacity
230 mg/g resin
PBI resin

Loading curve
for 3-HPA broth
on PBI resin

Bed volumes (BV) Effluent:

Influent:
Ethanol
B.

PBI resin
Effluent conc. (g/L)

PBI resin

Ethanol elution
curve for 3-HPA
broth on PBI
resin

Ethanol
Bed volumes (BV) effluent:
Figure S33. (A) Loading curve for acidified 3-HP broth from glucose finding a loading capacity of 230
mg/g dry resin for the polybenzimidazole (PBI) resin. Note that a capacity of at least 50 mg acid / g dry
resin is needed for the resin to be industrially relevant (99). (B) Elution curve of 3-HP from the resin using
ethanol as the eluent.

S52
FERMENTATION SIMULATED MOVING BED
DEWATERING
T = 78 °C
Ethanol
recycle
Ethanol
eluent
makeup
pH ~3
Column 1
Trays = 20
clarified 5 wt. % 3-HP
broth 3-HP
2 wt. % H2O

(aq.)
~ 10 wt. % 93 wt.% EtOH

cells +
debris Aqueous waste
To WWT
T = 93 °C

50 wt. % 3-HP
50 wt.% EtOH TO REACTIVE
Fermentation baseline model DISTILLATION
parameters
pH = 3 SMB baseline model parameters
Efficiency = 98%
3-HP yield from sugars = 80%
PBI Resin
Total sugar utilization = 98%

Figure S34. Graphical representation of the process flow diagram for the fermentation operation to
produce 3-HP at low pH and the dewatering operation used to produce a concentrated stream of 3-HP in
ethanol via a simulated moving bed (SMB) and distillation column.

S53
REACTIVE DISTILLATION
4-hydroxy TEMPO (100 ppm)

Ternary azeotrope H 2O
40 wt. % EtOH Column 2 overhead
10 wt,% H2O
50 wt.% ethyl acrylate Decanter
CSTR parameters
2-4 wt.% H2SO4 Aqueous to WWT
46-48 wt.% 3-HP
50 wt.% EtOH T = 40 °C
3 hr residence time H2SO4
90°C Organic reflux

Ethyl acrylate
50 wt. % 3-HP Ethanol
50 wt.% EtOH water Column 2 Column 3
Trays = 24 Trays = 7

Ethanol
H2SO4 Recycle
3-HP

98 wt%
ethyl
acrylate
TO
NITRILATION
Reactive distillation baseline model parameters
Conversion to ethyl acrylate = 100%
Efficiency = 98%

Figure S35. Graphical representation of the reactive distillation operation used in the economic analysis.
Design parameters were taken from previously published work available in references (25, 26).

S54
N2

H 2O Flowrates into reactor! NH3


N2: 1923 sccm (8.58E-2 mol/min)!
Ethyl acrylate: 0.077 ml/min (6.87E-4 mol/min)!
NH3: 128 sccm (0.0057 mol/min)!
EtOH !

!
Reactor parameters! Nitrilation
WHSV: 0.10hr-1! TiO2 310°C
mol / min

Temp: 310°C!
Contact time ~0.48 sec!
43g TiO2!
!

H 2O

EtOH
TiO2

TOS (hr)
Figure S36. Complete dataset and conditions used for the results shown in Figure 4 of the main text. The
graph displays all reaction products measure in molar units. Complete reaction conditions used are given
in the grey box.

S55
Ethyl Acrylate Nitrilation! ΔH 0rxn!
TiO2
+" NH3 310 °C
+" +""H2O +81.2 kJ/mol!

Figure S37. Reaction enthalpy for the nitrilation of ethyl acrylate to form acrylonitrile ethanol and water.
This reaction enthalpy was calculated from the values in Table S2. (+179.7+-234+-241.8)-(-331.4+-45.9)
= +81.2 kJ/mol.

S56
Increasing PNH3 PNH3
2.16 kPa

3.20 kPa

4.22 kPa
A/A0

8.11 kPa

2.16 kPa fit

TiO2 3.20 kPa fit


310°C 4.22 kPa fit
1.08 kPa ethyl acrylate
8.11 kPa fit
Varying NH3

Time (min)
Figure S38. Activity loss curves (A/A0) displaying the deactivation of the titania catalyst as a function of
time at 310°C. Four curves are present representing different partial pressures of ammonia while keeping
the partial pressure of ethyl acrylate constant at 1.08 kPa. These results indicate that increasing the
partial pressure of ammonia decreases the rate of catalyst deactivation. The fitted lines represent an
empirical fit using both this data set and that shown in Figure S46 discussed in the SI text.

S57
PEA
0.45 kPa

0.89 kPa

1.09 kPa
A/A0

Increasing PEA 1.55 kPa

0.45 kPa fit

TiO2 0.89 kPa fit


310°C 1.09 kPa fit
4.23 kPa NH3
1.55 kPa fit
Varying ethyl acrylate

Time (min)
Figure S39. Activity loss curves (A/A0) displaying the deactivation of the titania catalyst as a function of
time at 310°C. Four curves are present representing different partial pressures of ethyl acrylate while
keeping the partial pressure of ammonia constant at 4.23 kPa. These results indicate that increasing the
partial pressure of ethyl acrylate increases the rate of catalyst deactivation. The fitted lines represent an
empirical fit using both this data set and that shown in Figure S45 discussed in the SI text.

S58
g coke / g TiO2

A/A0
Figure S40: Correlation between the grams of coke burned off of the TiO2 catalyst as a function of activity
lost, assuming a linear fit.

S59
2 % yield loss
at 0.5 minutes

20 % yield loss
at 5 minutes

A/A0

TiO2
310°C
90.0 kPa NH3
11.26 kPa EA
Full scale design parameters

Time (min)
Figure S41: Activity loss curve calculated from the empirical model for activity loss (Equation S5) for the
full-scale process design equations. This result calculates that a 0.5 minute regeneration cycle time is
needed for the catalyst to maintain ≥ 98% yields of ACN from ethyl acrylate.

S60
NITRILATION
TiO2 ACN
310°C CO2 + H2O H2O
T = 78 °C EtOH
Entrained NH3 TO ACN
TiO2 SEPARATION
550°C TRAIN 1

DMSO
Ammonia
accumulator
Ethyl
acrylate NH3
Gas Column 4
stripper Trays = 10
NH3
makeup DMSO
NH3 ACN ACN
EtOH H2O
H2O EtOH
NH3 Entrained NH3
Air

Nitrilation baseline model


parameters T = 191 °C
Efficiency = 98%

Figure S42. Graphical representation of the process flow diagram used to model the nitrilation reaction
unit operations. As described in the SI, nitrilation is modeled to occur in a riser reactor system, where
catalyst regeneration occurs continuously. Reaction products are captured in a DMSO gas stripping tower
that contains 100 ppm of 4-hydroxy TEMPO as a polymerization inhibitor solved in the DMS. The DMSO
is then passed through a distillation column to recover the water, ACN, ethanol, and the DMSO and
inhibitor is recycled back to the stripping tower.

S61
ACN SEPARATION TRAIN 1
4-hydroxy TEMPO (100 ppm)
T = 70 °C
PURIFIED ACN
97%
Decanter
H2SO4
ACN
H2O In-line
EtOH mixer
Entrained NH3
Column 5
Trays = 45

T = 80 °C
H2O
Ethanol
Ammonium sulfate TO ACN
SEPARATION
TRAIN 2
ACN separation train (1 and 2 combined)
baseline model parameters
Efficiency = 97%

Figure S43. Graphical representation of the process flow diagram used to model the first section of the
ACN separation train where ACN is recovered at a purity of 97% from a distillation tower and decanter
system. The bottoms product from the column is passed to the section of the ACN separation train shown
in Figure S44.

S62
ACN SEPARATION TRAIN 2

Molecular
T = 78 °C EtOH Sieve drying
H2O EtOH
Recycle to
SMB

H2O
Ethanol
Ammonium sulfate
Column 6
Trays = 40

T = 100 °C

H2O
Ammonium sulfate
To WWT

ACN separation train (1 and 2 combined)


baseline model parameters
Efficiency = 97%

Figure S44. Graphical representation of the process flow diagram used to model the second section of
the ACN separation train. Here the bottoms product from the previous distillation column (Figure S43) is
fed to the middle of a second distillation column to recover ethanol which is dried using molecular sieves
and the ethanol product is recycled back to the SMB dewatering operation in Figure S34.

S63
Feedstock & Handling
Pretreatment & Conditioning
Hydrolysis & Bioconversion to 3-HP
ACN Production
Cellulase Enzyme Production
Wastewater Treatment
Heat Generation
Utilities

Figure S45. Cost breakdown for the major process areas in the production of ACN from lignocellulosic
biomass.

S64
Figure S46. Tornado plot summarizing the results of the sensitivity analysis performed on the process.

S65
Figure S47. Life cycle stages considered in the FTG pathway for biomass-derived ACN from 2G sugars.

S66
Figure S48. Life cycle stages considered in the 1G sugar-derived ACN.

S67
SUPPLEMENTARY TABLES

Table S1. Anatase TiO2 computed and experimental lattice parameters and volume.
3
Method a (Å) c (Å) V (Å ) Source
PBE-TS+U2.5 eV 3.826 9.565 140.0 this work
Sorescu et
PBE-TS+U2.5 eV 3.825 9.588 140.3
al.(53)
Experiment 3.782 9.502 135.9 Burdett et al.(56)

S68
Table S2. Ethyl acrylate nitrilation enthalpies of formation and reaction enthalpies in kJ/mol.

NIST DIPPR G4 PBE-TS+U


Enthalpies of formation (∆𝒇 𝑯°𝒈𝒂𝒔 , kJ/mol)
-354.2
ethyl acrylate (EA) -331.4 -339.3
-327.8
ethyl-3-hydroxypropanoate (E3HP) - -615.1
ethyl lactate (EL) - -695.1 -622.7
-45.9
NH3 -43.7
-45.94
Ethanol (EtOH) -234 -233.6
H 2O -241.83 -240.1
172.6
Acrylonitrile (ACN) 181.8
179.7
Enthalpies of reaction (∆𝑯°𝒓𝒙𝒏 , kJ/mol)
E3HP → EA + H2O 99 - 126 36 39
EA + NH3 → ACN + H2O + EtOH 70 - 104 91 116
E3HP + NH3 → ACN + EtOH + 2H2O - 127 154
EL + NH3 → ACN + EtOH + 2H2O 196 - 203 135 -

S69
Table S3. Reaction energetics associated with ethyl-3-hydroxypropanoate dehydration in kJ/mol (D) at 0
K. The quantities ΔE, EA,f, and EA,r represent the reaction energy and the barriers for the forward and the
reverse reaction, respectively. All values (ZPE-corrected in parentheses) are given in kJ/mol.
Abbreviations stand for ethyl-3-hydroxypropanoate (ethyl 3-HP), ethyl acrylate (EA), and an alkoxide
intermediate resulting from ethyl 3-HP dehydrogenation (ethyl 3-HP-H). Star symbol (*) denotes a
surface-bound species.

Reaction ΔE EA,f EA,r


Adsorption
D1 E3HP → E3HP* -138 - -
D2 EA → EA* -32 - -
D3 H2O → H2O* -89 - -
O-H Scission
D4 E3HP* → (E3HP-H)* + H* 15 - -
D5 OH* → O* + H* 41 - -
D6 H2O* → H* + OH* 22 (25) 49 (40) 27 (14)
C-O and C-H scission
D7 (E3HP-H)* + H* → EA* + 2H* + O* 111 - -
D8 E3HP* → EA* + OH* + H* 65 (51) 118 (103) 53 (52)

S70
Table S4. Reaction energetics for aminolysis (A) at 0 K. The quantities ΔE, EA,f, and EA,r represent the
reaction energy and the barriers for the forward and the reverse reaction, respectively. All values (ZPE-
corrected in parentheses) are given in kJ/mol. Abbreviations stand for ethyl acrylate (EA), ethanol (EtOH),
an ethoxide intermediate resulting from EtOH O-H scission (EtO), 1-amino-1-ethoxy-2-propen-1-ol
intermediate (AEP), acrylamide (AAm), and surface acrylate (Acr). The numbering for EA and EtOH
surface species (e.g., EAn) represents the various conformations along the respective paths. Star symbol
(*) denotes a surface-bound species.

Name Reaction ΔE (kJ/mol) EA,f EA,r


Adsorption
A1 EA → EA1* -147 (-142) - -
A1' EA → EA2* -121 (-115) - -
A1'' EA → EA4* -143 (-138) - -
A2 NH3 → NH3* -132 (-120) - -
A2' NH3 → NH3* (EA1* spectator) -117 (-105) - -
A3 AAm → AAm* -165 (-157) - -
A4 EtOH → EtOH1* -74 (-68) - -
A4' EtOH → EtOH2* -122 (-116) - -
NHx insertion
A5 EA1* + NH3* → AAm* + EtOH1* 36 (34) 270 (259) 234 (225)
A6 EA3* + NH2* + H* → AEP* -62 (-49) 38 (32) 101 (81)
A7 Acr* + NH2* → AAm* -85 (-79) 96 (98) 180 (177)
N-H scission
A8 NH3* → NH2* + H* 113 (104) 113 (104) 0 (0)
C-O scission
A9 AEP* → AAm* + EtO* + H* -40 (-49) 33 (19) 74 (67)
A10 EA5* → Acr* + EtO* -95 (-97) 70 (66) 165 (163)
O-H scission
A11 EtOH2* → EtO* + H* 17 (15) 57 (44) 40 (29)
Conformational change
A12 EA2* → EA3* 59 (58) 75 (72) 16 (14)
A13 EA4* → EA5* 88 (86) - -

S71
Table S5. Reaction energetics for acrylamide dehydration (AD) at 0 K in kJ/mol. The quantities ΔE, EA,f,
and EA,r represent the reaction energy and the barriers for the forward and the reverse reaction,
respectively. All values (ZPE-corrected in parentheses) are given in kJ/mol. Abbreviations stand for
acrylamide (AAm), enol form of acrylamide (Enol), ethanol (EtOH), an ethoxide intermediate resulting
from EtOH O-H scission (EtO), partially dehydrogenated acrylamide resulting from a single N-H scission
(AAm-H) and two N-H scission reactions (AAm-2H), partially dehydrogenated enol form of acrylamide
resulting from N-H scission (Enol-H) and acrylonitrile (ACN). The numbering for AAm surface species
(e.g., AAm1) represents the conformations along the respective reaction paths. Star symbol (*) denotes a
surface-bound species.

Name Reaction ΔE EA,f EA,r


Adsorption
AD1 AAm → AAm1* -165 (-157) - -
AD1' AAm → AAm* (EtO*,H* spectators) -127 (-121) - -
AD2 ACN → ACN* -88 (-81) - -
AD3 H2O → H2O* -89 - -
Conformational change
AD4 AAm1* → AAm2* 19 (19) - -
N-H scission & tautomerization
AD5 AAm2* à (AAm-H)* + H* 34 (31) 47 (40) 13 (9)
AD6 Enol* à (Enol-H)* + H* 72 (68) 104 (92) 32 (24)
AD7 (AAm-H)* + H* → (AAm-2H)* + 2H* 86
AD8 Enol* à ACN* + OH* + H* 53 (42) 140 (125) 87 (83)
AD9 (AAm-H)* + H* à ACN* + 2H* + O* 156
AD10 AAm* + EtO* + H* à Enol* + EtOH* 5 (6) 7 (-3) 6 (-8)
C-O/C-OH scission
AD11 (Enol-H)* + H* à ACN* + OH* + H* -19 (-26) 38 (32) 58 (59)
AD12 (AAm-2H)* + 2H* → ACN* + 2H* + O* 70
O-H scission
AD13 Enol* à (AAm-H)* + H* -14 (-15) 17 (8) 31 (23)
AD14 ACN* + H* + OH* à ACN* + 2H* + O* 89
AD15 H2O* à H* + OH* 22 (25) 49 (40) 27 (14)

S72
Table S6: Measured acid site density and BET surface area for the fresh, spent, and regenerated TiO2
catalyst used in the reactions shown in Figure 2 of the main text

Fresh TiO2 Spent TiO2 Regenerated TiO2


-1
Acid sites (μmol g ) 160 ---- 200
2 -1
BET (m g ) 49 27 50

S73
Table S7. Delivered feedstock composition assumed in the process design.

Component Composition(dry wt %)
Glucan 35.05
Xylan 19.53
Lignin 15.76
Ash 4.93
Acetate* 1.81
Protein 3.10
Extractives 14.65
Arabinan 2.38
Galactan 1.43
Mannan 0.60
Sucrose 0.77
Total structural carbohydrate 58.99
Total structural carbohydrate + sucrose 59.76
Moisture (bulk wt %) 20.0
*Represents acetate groups present in the hemicellulose polymer; converted to acetic acid in pretreatment.

S74
Table S8. Summary of key process parameters and yields utilized in the baseline process design.

Design Basis
Deacetylation
Deacetylation Solids Loading (wt%) 20%
Deacetylation temperature (°C) 90
Sodium Hydroxide Loading (mg/g dry biomass) 14
Solubilized Xylan (wt%) 1.6%
Solubilized Glucan (wt%) 0.4%
Solubilized Arabinan (wt%) 2.2%
Ash Removal (wt%) 70%
Solubilized Lignin (wt%) 61%
Solubilized Acetates (wt%) 96%
Enzymatic Hydrolysis + Hydrolysate Conditioning
Enzymatic Hydrolysis Solids Loading (wt%) 20%
Enzyme Loading mg/g 10
Combined Saccharification Time (d) 3.5
Hydrolysis Glucan to Glucose 90%
Glucan to Oligomer Glucose 4%
Glucan to Cellobiose 1.2%
Xylan to Xylose 90%
Hydrolysis Arabinan to Arabinose 85%
S/L Separation: Soluble sugar loss to solids 5%
Bioconversion (3-HP) Production
Productivity (g/L-hr) 2.0
% Conversion: Glucose → 3-HP with a theoretical yield of 2.00 mol
-1 80% [98%]
3-HP mol glucose [total glucose utilization]*
% Conversion: Xylose → 3-HP with a theoretical yield of 1.67 mol
-1 80% [98%]
3-HP mol xylose [total xylose utilization]*
% Conversion: Arabinose → 3-HP with a theoretical yield of 1.67
-1 80% [98%]
mol 3-HP mol arabinose [total arabinose utilization*
Overall Separations Efficiency 98%
Reactive Distillation
% Conversion: 3-HP → ethyl acrylate 100%
Temperature CSTR 90°C
Pressure 1 atm
Acrylonitrile Production
% Conversion: ethyl acrylate → ACN 98%
% Conversion: ethyl acrylate → Coke 2%
Temperature 310°C
Pressure 1 atm
Catalyst:ethyl acrylate loading 10:1
Ammonia: ethyl acrylate (molar) 8:1
* First number represents sugar conversion to desired product (3-HP), values in brackets
indicate total sugar utilization (including diversions to cell mass)

S75
Table S9. Capital cost breakout for process design.

Process Area Purchased Cost Installed Cost


Area 200: Pretreatment/Neutrallization $ 31,181,690 $ 47,364,034
Area 300: Enzymatic
$ 63,111,165.50 $ 116,277,622.67
hydrolysis/conditioning/bioconversion
Area 400: ACN Production $ 87,978,717.83 $ 162,094,236
Area 500: Enzyme Production $ 7,182,074 $ 12,268,047
Area 600: Wastewater $ 48,753,584 $ 67,187,897
Area 700: Storage $ 1,286,223 $ 2,350,532
Area 800: Boiler $ 41,452,551 $ 75,044,777
Area 900: Utilities $ 7,169,915 $ 12,352,273
$ 288,115,920 $ 494,939,419
Warehouse 4.0% of ISBL $ 13,520,158
Site Development 9.0% of ISBL $ 30,420,355
Additional Piping 4.5% of ISBL $ 15,210,177
Total Direct Costs (TDC) $ 554,090,109
Prorateable Expenses 10.0% of TDC $ 55,409,011
Field Expenses 10.0% of TDC $ 55,409,011
Home Office & Construction Fee 20.0% of TDC $ 110,818,022
Project Contingency 10.0% of TDC $ 55,409,011
Other Costs (Start-Up, Permits, etc.) 10.0% of TDC $ 55,409,011
Total Indirect Costs $ 332,454,065
Fixed Capital Investment (FCI) $ 886,544,174
Land $ 1,848,000
Working Capital 5.0% of FCI $ 44,327,209
Total Capital Investment (TCI) $ 932,719,382

Lang Factor (FCI/Purchased Equip Cost) 3.7

TCI per annual million pounds $ 2.75/MM lb

All costs are in 2014 Dollars

S76
Table S10. Financial assumptions and design basis.

Plant life 30 years


Plant throughput 2000 dry metric tons/day biomass
Cost year dollar 2014$s
Capacity Factor 90%
Discount rate 10%
General plant depreciation 200% declining balance (DB)
General plant recovery period 7 years
Steam plant depreciation 150% DB
Steam plant recovery period 20 years
Federal tax rate 35%
Financing 40% equity
Loan terms 10-year loan at 8% APR
Construction period 3 years
First 12 months’ expenditures 8%
Next 12 months’ expenditures 60%
Last 12 months’ expenditures 32%
Working capital 5% of fixed capital investment
Start-up time 6 months
Revenues during start-up 50%
Variable costs during start-up 75%
Fixed costs during start-up 100%

S77
Table S11. Summary of economics for integrated ACN production design using lignocellulosic biomass.

Overall economic details


MSP ($/lb) $0.89
Annual Product Yield (MM lb/yr) 339.2
Total Capital Investment (MM$s) 932
Total Variable Operating Cost (MM$/yr) 18
Total Fixed Operating Cost (MM$/yr) 22

S78
Table S12. Cost of raw materials.

Component Cost (2014$) Source


Feedstock (corn stover) $84.45/dry metric ton U.S. DOE MYPP @ 20% moisture (70)
Sulfuric acid, 93% $0.0539 /lb Basic Chemical of Omaha via Harris Group (70)
Ammonia $0.2696 /lb Terra Industries via Harris Group (70)
Corn steep liquor $0.0341 /lb Corn Products via Harris Group (70)
Diammonium phosphate $0.5930 /lb Ronas Chemicals via Harris Group (70)
Glucose $0.3487 /lb USDA ERS (70)
SO2 $0.1825 /lb Davis et al. (70)
Enzyme nutrients $0.4934 /lb Davis et al. (70)
Titanium Dioxide $9.07/lb Adopted from Dutta et al. (75)
Caustic $0.0899 /lb Input from subcontractor (22)
Required resin make-up $4MM /yr Input from subcontractor (22)
Filter replacements $5 MM/yr Input from subcontractor (22)
Polymer for WWT $2.65 /lb Input from subcontractor (22)
Lime $0.1198 /lb Input from subcontractor (22)
Boiler chemicals $3.01 /lb Aden et al. (100)
Cooling tower chemicals $1.798 /lb Aden et al. (100)
Fresh water $0.0002 /lb (76, 100)

S79
Table S13. Key inventory data reported per kg of bio-ACN produced from lignocellulosic sugars.

Resource Consumption
Corn Stover (dry biomass) 4.27E+00 kg
Sulfuric Acid (H2SO4) 7.24E-02 kg
Sodium Hydroxide (NaOH) 5.98E-02 kg
Ammonia (NH3) 3.60E-01 kg
Corn Steep Liquor 6.45E-02 kg
Diammonium Phosphate ((NH4)2HPO4) 9.00E-03 kg
Flocculants 1.47E-02 kg
Glucose 6.89E-02 kg
Host Nutrients 1.92E-03 kg
Sulfur Dioxide (SO2) 4.68E-04 kg
Dimethyl Sulfoxide ((CH3)2SO) 2.03E-02 kg
Wastewater Treatment Chemicals 1.47E-04 kg
Lime (CaO) 8.44E-03 kg
Cooling Tower Chemicals 2.72E-04 kg
Makeup Water 1.53E+01 kg
Titanium Dioxide (TiO2) Catalyst 4.87E-03 kg
Natural Gas 0.00E+00 kg
Grid Electricity 1.20E+00 kWh
Solid and Liquid Wastes
Ash Disposal 2.31E-01 kg
Wastewater 4.62E-01 kg
Direct Air Emissions
Water (H2O) 2.01E+01 kg
Nitrogen (N2) 2.08E+01 kg
Oxygen (O2) 2.60E+00 kg
Fossil CO2 0.00E+00 kg
Biogenic CO2 4.60E+00 kg
Methane (CH4) 1.23E-04 kg
Nitrogen dioxide (NO2) 4.16E-03 kg
Carbon monoxide (CO) 3.86E-03 kg
Sulfur dioxide (SO2) 5.07E-04 kg

S80
Table S14: Key inventory data reported per kg of bio-ACN produced from sugars from Number 11
commodity (raw) sugar (1G)

Resource Consumption
Feedstock - Number 11 commodity (raw) sugar 2.30E+00 kg
Sulfuric Acid (H2SO4) 5.87E-02 kg
Sodium Hydroxide (NaOH) 0.00E+00 kg
Ammonia (NH3) 3.44E-01 kg
Dimethyl Sulfoxide ((CH3)2SO) 2.03E-02 kg
Wastewater Treatment Chemicals 2.10E-05 kg
Lime (CaO) 1.67E-03 kg
Cooling Tower Chemicals 1.94E-04 kg
Makeup Water 2.08E+01 kg
Titanium Dioxide (TiO2) Catalyst 4.87E-03 kg
Natural Gas 5.12E-01 kg
Grid Electricity 3.52E-01 kWh
Ethanol Replacement 2.52E-02 kg
Solid and Liquid Wastes
Ash Disposal 4.13E-03 kg
Wastewater 3.66E-01 kg
Direct Air Emissions
Water (H2O) 1.46E+01 kg
Nitrogen (N2) 1.31E+01 kg
Oxygen (O2) 1.12E+00 kg
Fossil CO2 3.25E+00 kg
Biogenic CO2 2.28E+00 kg
Methane (CH4) 3.52E-05 kg
Nitrogen dioxide (NO2) 7.66E-04 kg
Carbon monoxide (CO) 7.66E-04 kg
Sulfur dioxide (SO2) 1.01E-04 kg

S81
Table S15: Acrylonitrile production life cycle GHGs comparison.

Number 11
commodity
(raw) sugar Lignocellulosic Petroleum-
Feedstock Source (1G) Sugars (2G) derived
GHG, kg CO2e/kg Acrylonitrile (C3H3N) 3.02 2.87 3.34
GHG Reduction 9.58% 14.1% --

S82
References
1. J. M. Thomas, W. J. Thomas, Principles and Practice of Heterogeneous Catalysis (Wiley,
2015).
2. R. K. Grasselli, F. Trifirò, Acrylonitrile from biomass: Still far from being a sustainable
process. Top. Catal. 59, 1651–1658 (2016). doi:10.1007/s11244-016-0679-7
3. The Sohio Acrylonitrile Process (American Chemical Society, 1996);
www.acs.org/content/dam/acsorg/education/whatischemistry/landmarks/acrylonitrile/sohi
o-acrylonitrile-process-commemorative-booklet-1996.pdf.
4. R. K. Grasselli, Advances and future trends in selective oxidation and ammoxidation catalysis.
Catal. Today 49, 141–153 (1999). doi:10.1016/S0920-5861(98)00418-0
5. R. K. Grasselli, in Nanostructured Catalysts: Selective Oxidations, C. Hess, R. Schlögl, Eds.
(Royal Society of Chemistry, 2011), pp. 96–140.
6. S. J. Park, Carbon Fibers (Springer, 2014).
7. G. Centi, S. Perathoner, F. Trifirò, V-Sb-oxide catalysts for the ammoxidation of propane.
Appl. Catal. A 157, 143–172 (1997). doi:10.1016/S0926-860X(97)00013-6
8. J. L. Dubois, Method for the synthesis of acrylonitrile from glycerol. U.S. Patent
CN101636381 (2010).
9. M. O. Guerrero-Pérez, M. A. Bañares, New reaction: Conversion of glycerol into acrylonitrile.
ChemSusChem 1, 511–513 (2008). doi:10.1002/cssc.200800023 Medline
10. C. Liebig, S. Paul, B. Katryniok, C. Guillon, J.-L. Couturier, J.-L. Dubois, F. Dumeignil, W.
F. Hoelderich, Glycerol conversion to acrylonitrile by consecutive dehydration over
WO3/TiO2 and ammoxidation over Sb-(Fe,V)-O. Appl. Catal. B 132–133, 170–182
(2013). doi:10.1016/j.apcatb.2012.11.035
11. J. Le Nôtre, E. L. Scott, M. C. R. Franssen, J. P. M. Sanders, Biobased synthesis of
acrylonitrile from glutamic acid. Green Chem. 13, 807–809 (2011).
doi:10.1039/c0gc00805b
12. G. D. Epps, E. E. Reid, Studies in the preparation of nitriles. III. The catalytic preparation of
nitriles. J. Am. Chem. Soc. 38, 2128–2135 (1916). doi:10.1021/ja02267a023
13. D. Decoster, S. Hoyt, S. Roach, Dehydration of 3-hydroxypropionic acid to acrylic acid.
Patent WO2013192451, USA, 2013.
14. A. C. Stevenson, Ammonolysis. Ind. Eng. Chem. 43, 1920–1924 (1951).
doi:10.1021/ie50501a012
15. S. Itagaki, K. Kamata, K. Yamaguchi, N. Mizuno, A monovacant lacunary silicotungstate as
an efficient heterogeneous catalyst for dehydration of primary amides to nitriles.
ChemCatChem 5, 1725–1728 (2013). doi:10.1002/cctc.201300063
16. B. V. Suvorov, N. R. Bukeikhanov, L. V. Li, A. Z. Zulkasheva, Ammonolysis of esters of
hydroxybenzoic acids on a boron phosphate catalyst. J. Appl. Chem. USSR 60, 677–679
(1987).
17. S. M. Pasternak, E. V. Pivovarova, I. Y. Lubyanitskii, Products of ammonolysis of dimethyl
esters of aliphatic dicarboxylic acids. J. Appl. Chem. USSR 47, 2590–2592 (1973).
18. A. Mekki-Berrada, S. Bennici, J. P. Gillet, J. L. Couturier, J. L. Dubois, A. Auroux, Fatty
acid methyl esters into nitriles: Acid-base properties for enhanced catalysts. J. Catal. 306,
30–37 (2013). doi:10.1016/j.jcat.2013.05.032
19. P. Kostestkyy, J. Yu, R. J. Gorte, G. Mpourmpakis, Structure-activity relationships on metal-
oxides: Alcohol dehydration. Catal. Sci. Technol. 4, 3861–3869 (2014).
doi:10.1039/C4CY00632A
20. C. Rathnasingh, S. M. Raj, Y. Lee, C. Catherine, S. Ashok, S. Park, Production of 3-
hydroxypropionic acid via malonyl-CoA pathway using recombinant Escherichia coli
strains. J. Biotechnol. 157, 633–640 (2012). doi:10.1016/j.jbiotec.2011.06.008 Medline
21. I. Borodina, K. R. Kildegaard, N. B. Jensen, T. H. Blicher, J. Maury, S. Sherstyk, K.
Schneider, P. Lamosa, M. J. Herrgård, I. Rosenstand, F. Öberg, J. Forster, J. Nielsen,
Establishing a synthetic pathway for high-level production of 3-hydroxypropionic acid in
Saccharomyces cerevisiae via β-alanine. Metab. Eng. 27, 57–64 (2015).
doi:10.1016/j.ymben.2014.10.003 Medline
22. M. J. Biddy, R. Davis, D. Humbird, L. Tao, N. Dowe, M. T. Guarnieri, J. G. Linger, E. M.
Karp, D. Salvachúa, D. R. Vardon, G. T. Beckham, The techno-economic basis for
coproduct manufacturing to enable hydrocarbon fuel production from lignocellulosic
biomass. ACS Sustain. Chem. Eng. 4, 3196–3211 (2016).
doi:10.1021/acssuschemeng.6b00243
23. S. Choi, C. W. Song, J. H. Shin, S. Y. Lee, Biorefineries for the production of top building
block chemicals and their derivatives. Metab. Eng. 28, 223–239 (2015).
doi:10.1016/j.ymben.2014.12.007 Medline
24. M. Grendze, F. Verhoff, Thermally-managed separation and dewatering processes for
recovering acid products. U.S. Patent 6146534 (2000).
25. F. W. Shaver, Method of producing alkyl esters of alpha-beta unsaturated monocarboxylic
acids froma beta-lactone and an alcohol. U.S. Patent 2510423 (1950).
26. I. L. Chien, K. Chen, C.-L. Kuo, Overall control strategy of a coupled reactor/columns
process for the production of ethyl acrylate. J. Process Contr. 18, 215–231 (2008).
doi:10.1016/j.jprocont.2007.02.006
27. J. C. Kelly, J. L. Sullivan, A. Burnham, A. Elgowainy, Impacts of vehicle weight reduction
via material substitution on life-cycle greenhouse gas emissions. Environ. Sci. Technol.
49, 12535–12542 (2015). doi:10.1021/acs.est.5b03192 Medline
28. S. Pacala, R. Socolow, Stabilization wedges: Solving the climate problem for the next 50
years with current technologies. Science 305, 968–972 (2004).
doi:10.1126/science.1100103 Medline
29. Tecnon OrbiChem, Chem-Net Facts: Acrylonitrile (2013);
www.orbichem.co.uk/userfiles/CNF%20Samples/acn_13_11.pdf.
30. Y. Li, X. Wang, X. Ge, P. Tian, High production of 3-hydroxypropionic acid in Klebsiella
pneumoniae by systematic optimization of glycerol metabolism. Sci. Rep. 6, 26932
(2016). doi:10.1038/srep26932 Medline
31. H. S. Chu, Y. S. Kim, C. M. Lee, J. H. Lee, W. S. Jung, J.-H. Ahn, S. H. Song, I. S. Choi, K.
M. Cho, Metabolic engineering of 3-hydroxypropionic acid biosynthesis in Escherichia
coli. Biotechnol. Bioeng. 112, 356–364 (2015). doi:10.1002/bit.25444 Medline
32. T. Werpy, G. Petersen, Top Value Added Chemicals from Biomass: Volume I—Results of
Screening for Potential Candidates from Sugars and Synthesis Gas (National Renewable
Energy Laboratory, 2004); www.osti.gov/scitech/servlets/purl/15008859.
33. M. J. Biddy, C. Scarlata, C. Kinchin, Chemicals from Biomass: A Market Assessment of
Bioproducts with Near-Term Potential (National Renewable Energy Laboratory, 2016);
www.nrel.gov/docs/fy16osti/65509.pdf.
34. K. Sundmacher, A. Kienle, Reactive Distillation (Wiley-VCH, 2002).
35. A. Orjuela, A. J. Yanez, L. Peereboom, C. T. Lira, D. J. Miller, A novel process for recovery
of fermentation-derived succinic acid. Separ. Purif. Tech. 83, 31–37 (2011).
doi:10.1016/j.seppur.2011.08.010
36. A. Chauvel, G. Lefebvre, Petrochemical Processes: Technical and Economic Characteristics
(Editions Technip, 1989).
37. J. McCallion, New separation approach saves catalyst and energy. Chem. Proc. 59, 73–74
(1996).
38. M. Sawa, M. Niwa, Y. Murakami, Relationship between acid amount and framework
aluminum content in mordenite. Zeolites 10, 532–538 (1990). doi:10.1016/S0144-
2449(05)80308-2
39. C. A. Emeis, Determination of integrated molar extinction coefficients for infrared
absorption bands of pyridine adsorbed on solid acid catalysts. J. Catal. 141, 347–354
(1993). doi:10.1006/jcat.1993.1145
40. C. P. Nash, A. Ramanathan, D. A. Ruddy, M. Behl, E. Gjersing, M. Griffin, H. Zhu, B.
Subramaniam, J. A. Schaidle, J. E. Hensley, Mixed alcohol dehydration over Brønsted
and Lewis acidic catalysts. Appl. Catal. A 510, 110–124 (2016).
doi:10.1016/j.apcata.2015.11.019
41. M. R. Basila, T. R. Kantner, K. H. Rhee, The nature of the acidic sites on a silica-alumina.
Characterization by infrared spectroscopic studies of trimethylamine and pyridine
chemisorption. J. Phys. Chem. 68, 3197–3207 (1964). doi:10.1021/j100793a020
42. E. P. Parry, An infrared study of pyridine adsorbed on acidic solids. Characterization of
surface acidity. J. Catal. 2, 371–379 (1963). doi:10.1016/0021-9517(63)90102-7
43. M. I. Zaki, M. A. Hasan, F. A. Al-Sagheer, L. Pasupulety, In situ FTIR spectra of pyridine
adsorbed on SiO2–Al2O3, TiO2, ZrO2 and CeO2: General considerations for the
identification of acid sites on surfaces of finely divided metal oxides. Colloids Surf. A
190, 261–274 (2001). doi:10.1016/S0927-7757(01)00690-2
44. G. Kresse, J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations
using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 (1996).
doi:10.1103/PhysRevB.54.11169 Medline
45. G. Kresse, J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and
semiconductors using a plane-wave basis set. Comput. Mater. Sci. 6, 15–50 (1996).
doi:10.1016/0927-0256(96)00008-0
46. G. Kresse, J. Hafner, Ab initio molecular dynamics for liquid metals. Phys. Rev. B 47, 558–
561 (1993). doi:10.1103/PhysRevB.47.558 Medline
47. G. Kresse, J. Hafner, Ab initio molecular-dynamics simulation of the liquid-metal-
amorphous-semiconductor transition in germanium. Phys. Rev. B 49, 14251–14269
(1994). doi:10.1103/PhysRevB.49.14251 Medline
48. G. Kresse, D. Joubert, From ultrasoft pseudopotentials to the projector augmented-wave
method. Phys. Rev. B 59, 1758–1775 (1999). doi:10.1103/PhysRevB.59.1758
49. P. E. Blöchl, Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994).
doi:10.1103/PhysRevB.50.17953 Medline
50. J. P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made simple.
Phys. Rev. Lett. 77, 3865–3868 (1996). doi:10.1103/PhysRevLett.77.3865 Medline
51. A. Tkatchenko, M. Scheffler, Accurate molecular van der Waals interactions from ground-
state electron density and free-atom reference data. Phys. Rev. Lett. 102, 073005 (2009).
doi:10.1103/PhysRevLett.102.073005 Medline
52. S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, A. P. Sutton, Electron-
energy-loss spectra and the structural stability of nickel oxide: An LSDA+U study. Phys.
Rev. B 57, 1505–1509 (1998). doi:10.1103/PhysRevB.57.1505
53. D. C. Sorescu, S. Civiš, K. D. Jordan, Mechanism of oxygen exchange between CO2 and
TiO2(101) anatase. J. Phys. Chem. C 118, 1628–1639 (2014). doi:10.1021/jp410420e
54. F. D. Murnaghan, The compressibility of media under extreme pressures. Proc. Natl. Acad.
Sci. U.S.A. 30, 244–247 (1944). doi:10.1073/pnas.30.9.244 Medline
55. F. Birch, Finite elastic strain of cubic crystals. Phys. Rev. 71, 809–824 (1947).
doi:10.1103/PhysRev.71.809
56. J. K. Burdett, T. Hughbanks, G. J. Miller, J. W. Richardson, J. V. Smith, Structural-electronic
relationships in inorganic solids: Powder neutron diffraction studies of the rutile and
anatase polymorphs of titanium dioxide at 15 and 295 K. J. Am. Chem. Soc. 109, 3639–
3646 (1987). doi:10.1021/ja00246a021
57. G. Makov, M. C. Payne, Periodic boundary conditions in ab initio calculations. Phys. Rev. B
51, 4014–4022 (1995). doi:10.1103/PhysRevB.51.4014 Medline
58. J. Neugebauer, M. Scheffler, Adsorbate-substrate and adsorbate-adsorbate interactions of Na
and K adlayers on Al(111). Phys. Rev. B 46, 16067–16080 (1992).
doi:10.1103/PhysRevB.46.16067 Medline
59. G. Henkelman, B. P. Uberuaga, H. Jónsson, A climbing image nudged elastic band method
for finding saddle points and minimum energy paths. J. Chem. Phys. 113, 9901–9904
(2000). doi:10.1063/1.1329672
60. Gaussian 09 (Wallingford CT, 2009).
61. L. A. Curtiss, P. C. Redfern, K. Raghavachari, Gaussian-4 theory. J. Chem. Phys. 126,
084108 (2007). doi:10.1063/1.2436888 Medline
62. K. A. Datsenko, B. L. Wanner, One-step inactivation of chromosomal genes in Escherichia
coli K-12 using PCR products. Proc. Natl. Acad. Sci. U.S.A. 97, 6640–6645 (2000).
doi:10.1073/pnas.120163297 Medline
63. M. D. Lynch, R. T. Gill, T. E. Lipscomb, Methods for producing 3-hydroxypropionic acid
and other products. U.S. Patent 8,883,464 (2014).
64. X. Chen, J. Shekiro, M. A. Franden, W. Wang, M. Zhang, E. Kuhn, D. K. Johnson, M. P.
Tucker, The impacts of deacetylation prior to dilute acid pretreatment on the bioethanol
process. Biotechnol. Biofuels 5, 8 (2012). doi:10.1186/1754-6834-5-8 Medline
65. H. Kwak, D. W. Hwang, Y. K. Hwang, J.-S. Chang, Recovery of alkyl lactate from
ammonium lactate by an advanced precipitation process. Separ. Purif. Tech. 93, 25–32
(2012). doi:10.1016/j.seppur.2012.03.025
66. L. A. Alves, J. B. Almeida e Silva, M. Giulietti, Solubility of D-glucose in water and
ethanol/water mixtures. J. Chem. Eng. Data 52, 2166–2170 (2007).
doi:10.1021/je700177n
67. D. Pedersen, C. Rosenbohm, Dry column vacuum chromatography. Synthesis 2001, 2431–
2434 (2001). doi:10.1055/s-2001-18722
68. N. Kawabata, J.-i. Yoshida, Y. Tanigawa, Removal and recovery of organic pollutants from
aquatic environment. 4. Separation of carboxylic acids from aqueous solution using
crosslinked poly(4-vinylpyridine). Ind. Eng. Chem. Prod. Res. Dev. 20, 386–390 (1981).
doi:10.1021/i300002a030
69. L. Tao, D. Schell, E. Tan, R. Elander, Achievement of Ethanol Cost Targets: Biochemical
Ethanol Fermentation via Dilute-Acid Pretreatment and Enzymatic Hydrolysis of Corn
Stover (National Renewable Energy Laboratory, 2014);
www.nrel.gov/docs/fy14osti/61563.pdf.
70. R. Davis et al., Process Design and Economics for the Conversion of Lignocellulosic
Biomass to Hydrocarbons: Dilute-Acid and Enzymatic Deconstruction of Biomass to
Sugars and Biological Conversion of Sugars to Hydrocarbons (National Renewable
Energy Laboratory, 2014); www.nrel.gov/docs/fy14osti/60223.pdf.
71. V. Kumar, S. Ashok, S. Park, Recent advances in biological production of 3-
hydroxypropionic acid. Biotechnol. Adv. 31, 945–961 (2013).
doi:10.1016/j.biotechadv.2013.02.008 Medline
72. C. S. Henry, L. J. Broadbelt, V. Hatzimanikatis, Discovery and analysis of novel metabolic
pathways for the biosynthesis of industrial chemicals: 3-hydroxypropanoate. Biotechnol.
Bioeng. 106, 462–473 (2010). Medline
73. K. R. Kildegaard, Z. Wang, Y. Chen, J. Nielsen, I. Borodina, Production of 3-
hydroxypropionic acid from glucose and xylose by metabolically engineered
Saccharomyces cerevisiae. Metab. Eng. Commun. 2, 132–136 (2015).
doi:10.1016/j.meteno.2015.10.001
74. S. Ramaswamy, H. J. Huang, B. V. Ramarao, Separation and Purification Technologies in
Biorefineries (Wiley, 2013).
75. A. Dutta, J. A. Schaidle, D. Humbird, F. G. Baddour, A. Sahir, Conceptual process design
and techno-economic assessment of ex situ catalytic fast pyrolysis of biomass: A fixed
bed reactor implementation scenario for future feasibility. Top. Catal. 59, 2–18 (2016).
doi:10.1007/s11244-015-0500-z
76. D. Humbird et al., Process Design and Economics for Biochemical Conversion of
Lignocellulosic Biomass to Ethanol (National Renewable Energy Laboratory, 2011);
www.nrel.gov/docs/fy11osti/47764.pdf.
77. R. Davis et al., Process Design and Economics for the Conversion of Lignocellulosic
Biomass to Hydrocarbons: Dilute-Acid and Enzymatic Deconstruction of Biomass to
Sugars and Catalytic Conversion of Sugars to Hydrocarbons (National Renewable
Energy Laboratory, 2015); www.nrel.gov/docs/fy15osti/62498.pdf.
78. R. Davis et al., Process Design and Economics for the Conversion of Lignocellulosic
Biomass to Hydrocarbons: Dilute-Acid Prehydrolysis and Enzymatic Hydrolysis
Deconstruction of Biomass to Sugars and Biological Conversion of Sugars to
Hydrocarbons (National Renewable Energy Laboratory, 2013);
www.nrel.gov/docs/fy14osti/60223.pdf.
79. E. C. D. Tan et al., Process Design and Economics for the Conversion of Lignocellulosic
Biomass to Hydrocarbons via Indirect Liquefaction: Thermochemical Research Pathway
to High-Octane Gasoline Blendstock Through Methanol/Dimethyl Ether Intermediates
(National Renewable Energy Laboratory, 2015); www.nrel.gov/docs/fy15osti/62402.pdf.
80. A. Dutta et al., Process Design and Economics for the Conversion of Lignocellulosic
Biomass to Hydrocarbons via Indirect Liquefaction: Thermochemical Research Pathway
to High-Octane Gasoline Blendstock Through Methanol/Dimethyl Ether Intermediates
(National Renewable Energy Laboratory, 2015); www.osti.gov/scitech/biblio/1015885.
81. S. B. Jones et al., Process Design and Economics for the Conversion of Lignocellulosic
Biomass to Hydrocarbon Fuels: Fast Pyrolysis and Hydrotreating Bio-Oil Pathway
(Pacific Northwest National Laboratory, 2013); www.nrel.gov/docs/fy14osti/61178.pdf.
82. Index Mundi, Commodity Sugar Prices;
www.indexmundi.com/commodities/?commodity=sugar.
83. W. L. Marques, V. Raghavendran, B. U. Stambuk, A. K. Gombert, Sucrose and
Saccharomyces cerevisiae: A relationship most sweet. FEMS Yeast Res. 16, fov107
(2016). doi:10.1093/femsyr/fov107 Medline
84. SimaPro Life Cycle Assessment Software (Product Ecology Consultants, Amersfoort,
Netherlands, 2014).
85. The Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET)
Model; https://greet.es.anl.gov/publications.
86. National Renewable Energy Laboratory, U.S. Life Cycle Inventory Database (2012);
www.nrel.gov/lci/.
87. R. Frischknecht, N. Jungbluth, H.-J. Althaus, G. Doka, R. Dones, T. Heck, S. Hellweg, R.
Hischier, T. Nemecek, G. Rebitzer, M. Spielmann, The ecoinvent database: Overview
and methodological framework. Int. J. Life Cycle Assess. 10, 3–9 (2005).
doi:10.1065/lca2004.10.181.1
88. B. G. Hermann, K. Blok, M. K. Patel, Producing bio-based bulk chemicals using industrial
biotechnology saves energy and combats climate change. Environ. Sci. Technol. 41,
7915–7921 (2007). doi:10.1021/es062559q Medline
89. C. S. M. Pereira, S. P. Pinho, V. M. T. M. Silva, A. E. Rodrigues, Thermodynamic
equilibrium and reaction kinetics for the esterification of lactic acid with ethanol
catalyzed by acid ion-exchange resin. Ind. Eng. Chem. Res. 47, 1453–1463 (2008).
doi:10.1021/ie071220p
90. H. Zhang, D. Chen, G. Zhang, S. Mi, N. Lu, Mechanism of the aminolysis of dimethyl
phenylphosphinate: A DFT study. J. Mol. Struct. Theochem 908, 12–18 (2009).
doi:10.1016/j.theochem.2009.04.039
91. R. Chen, X. Luo, G. Liang, Theoretical studies on the aminolysis mechanism of propylene
carbonate with ammonia. Theor. Chem. Acc. 134, 32 (2015). doi:10.1007/s00214-015-
1634-6
92. A. A. Tsyganenko, D. V. Pozdnyakov, V. N. Filimonov, Infrared study of surface species
arising from ammonia adsorption on oxide surfaces. J. Mol. Struct. 29, 299–318 (1975).
doi:10.1016/0022-2860(75)85038-1
93. R. A. Rajadhyaksha, G. W. Joshi, Dehydration of carboxamides to nitriles using sulphated
zirconia catalyst. Stud. Surf. Sci. Catal. 59, 479–486 (1991).
94. F. Duvernay, P. Chatron-Michaud, F. Borget, D. M. Birney, T. Chiavassa, Photochemical
dehydration of acetamide in a cryogenic matrix. Phys. Chem. Chem. Phys. 9, 1099–1106
(2007). doi:10.1039/b617586d Medline
95. R. B. Licht, A. T. Bell, A DFT investigation of the mechanism of propene ammoxidation
over α-bismuth molybdate. ACS Catal. 7, 161–176 (2016). doi:10.1021/acscatal.6b02523
96. E. L. Tollefson, R. M. Decker, C. B. Johnson, Development of a process for production of
acetonitrile from acetic acid and ammonia. Can. J. Chem. Eng. 48, 219–223 (1970).
doi:10.1002/cjce.5450480223
97. W. Ratchford, C. Rehberg, C. Fisher, Preparation of acrylic and methacrylic acids by
pyrolysis of their alkyl esters. J. Am. Chem. Soc. 66, 1864–1866 (1944).
doi:10.1021/ja01239a016
98. A. W. Schnizer, E. N. Wheeler, Condensation of acrylic acid. U.S. Patent 3462484 (1969).
99. B. H. Davison, N. P. Nghiem, G. L. Richardson, Succinic acid adsorption from fermentation
broth and regeneration. Appl. Biochem. Biotechnol. 113-116, 653–669 (2004).
doi:10.1385/ABAB:114:1-3:653 Medline
100. A. Aden et al., Lignocellulosic Biomass to Ethanol Process Design and Economics
Utilizing Co-current Dilute Acid Prehydrolysis and Enzymatic Hydrolysis for Corn
Stover (National Renewable Energy Laboratory, 2002);
www.nrel.gov/docs/fy02osti/32438.pdf.

You might also like