Supporting Information

A mechanistic model of fast pyrolysis of hemicellulose

Xiaowei Zhou,† Wenjun Li,‡ Ross Mabon,‡ and Linda J. Broadbelt*, †

† Department of Chemical and Biological Engineering, Northwestern University, 2145 Sheridan

Road, Evanston, IL 60208 United States.

‡ Corporate Strategic Research, ExxonMobil Research and Engineering Company, 1545 US Route

22 East, Annandale, NJ 08801, USA

*Correspondence concerning this article should be addressed to L. J. Broadbelt at

broadbelt@northwestern.edu; fax: +1-847-491-3728; tel: +1-847-491-5351.

S1

Electronic Supplementary Material (ESI) for Energy & Environmental Science.This journal is © The Royal Society of Chemistry 2018

Table S1. Comparison of model yields with experimental results of fast pyrolysis of extracted

cornstover hemicellulose at 500 °C

Model yields, wt%Products detected from

hemicellulose pyrolysis

Expt. Yields,1 wt% DP=15 DP=30 DP=45 DP=75 DP=135

Formaldehyde 0.25 ± 0.02 1.71 1.65 1.65 1.69 1.58

Acetaldehyde 1.15 ± 0.07 0.94 0.89 0.88 0.88 0.83

Furan 0.10 ± 0.00

— — — — —

Acetone 0.13 ± 0.02 — — — — —

Methyl glyoxal 3.31 ± 0.31 6.02 5.28 5.22 5.05 5.01

2-Methyl furan 0.09 ± 0.00 — — — — —

Glycolaldehyde 12.85 ± 1.09

14.85 13.96 13.97 14.10 13.42

Acetic acid 0.18 ± 0.02 — — — — —

Acetol 1.2 ± 0.04 2.39 2.20 2.39 2.28 2.29

2-Furaldehyde (C5H4O2) 2.2 ± 0.12 2.13 2.01 2.00 2.01 1.88

2-Furan methanol (C5H6O2) 0.31 ± 0.02 — — — — —

3-Furan methanol (C5H6O2) 0.17 ± 0.01 — — — — —

Other DAXP 1 0.49 ± 0.02 0.54 0.50 0.50 0.49 0.46

5-Methyl furfural (C6H6O2) 0.64 ± 0.04 — — — — —

1,2;4,5-Dianhydroxylopyranose (DAXP 1)

2.83 ± 0.23 2.93 2.62 2.63 2.60 2.64

2(5H)-Furanone 0.49 ± 0.04 — — — — —

1,2;3,4-Dianhydroxylopyranose (DAXP 2)

13.34 ± 0.20 12.40 15.28 14.51 14.55 13.93

S2

5-Hydroxymethylfurfural 0.20 ± 0.00 — — — — —

Other DAXP 2 0.88 ± 0.07 2.69 2.24 2.23 2.13 2.11

1,2-Anhydroxylopyranose (AXP) 7.14 ± 0.62 8.68 7.54 7.58 7.50 7.89

Levoglucosan-pyranose 0.81 ± 0.03 — — — — —

Levoglucosan-furanose 0.13 ± 0.00 — — — — —

Char 9.44 ± 1.19 9.11 8.81 9.15 8.93 8.87

CO 1.72 ± 0.29 2.18 2.11 2.19 2.14 2.11

CO2 6.02 ± 0.59 5.99 5.79 6.02 5.88 5.82

Water (calculated) 14.98 ± 1.00 15.57 15.30 14.99 14.33 14.30

Total 81.25 ± 2.75 88.27 86.42 86.20 84.97 82.55

Note that experimental yields of pyrolysis products and corresponding standard deviation for

triplicate pyrolysis runs are from experimental study of Zhang et al.1 In their experiments, only

carbohydrate derived products are analyzed, and the experimental yields of products were

normalized based on carbohydrate content.

S3

Table S2. Comparison of model yields for fast pyrolysis of extracted cornstover hemicellulose

with varying initial PDI (case of DP=75 at 500 °C)

Model yields, wt%,Hemicellulose pyrolysis products Expt yields,1

wt% PDI=1 PDI=1.06 PDI=1.5 PDI=2 PDI=2.5

Acetaldehyde 1.15 0.89 0.88 0.90 0.82 0.79

Methyl glyoxal 3.31 5.06 5.05 5.56 4.81 4.68

Glycolaldehyde 12.85 14.14 14.10 13.87 13.03 12.54

Acetol 1.2 2.28 2.28 2.52 2.24 2.21

2-Furaldehyde 2.2 2.13 2.13 2.22 2.03 1.98

1,2;4,5-DAXP 2.83 2.60 2.60 2.55 2.43 2.35

1,2;3,4-DAXP 13.34 14.49 14.55 15.36 15.71 16.00

1,2-AXP 7.14 7.51 7.50 7.16 6.99 6.75

Char 9.44 8.93 8.93 9.39 8.87 8.78

CO 1.72 2.14 2.14 2.25 2.12 2.10

CO2 6.02 5.88 5.88 6.18 5.83 5.77

Water (calculated) 14.98 14.28 14.33 15.14 14.93 15.12

S4

Table S3: Comparison of model yields with experimental results of fast pyrolysis of extracted

corn stover hemicellulose based on models that handle differently the decomposition of chain

species into monomers, dimers, and/or trimers.

Model yields, wt%,Major products of

hemicellulose pyrolysis

Expt yields,1

wt% Monomeric

DP=45

Dimeric

DP=45

Trimeric

DP=45

Trimeric

DP=135

Acetaldehyde 1.15 0.86 0.82 0.83 0.83

Methyl glyoxal 3.31 5.13 5.23 4.72 5.01

Glycolaldehyde 12.85 13.68 13.30 13.00 13.42

Acetol 1.2 1.26 1.30 1.31 1.32

2-Furaldehyde 2.2 1.96 1.86 1.86 1.88

1,2;4,5-DAXP 2.83 2.59 2.70 2.35 2.64

1,2;3,4-DAXP 13.34 14.93 13.65 16.21 13.93

1,2-AXP 7.14 7.51 8.42 6.60 7.89

Char 9.44 8.87 9.18 8.50 8.87

CO 1.72 2.12 2.20 2.03 2.11

CO2 6.02 5.83 6.04 5.59 5.82

Water (calculated) 14.98 14.88 15.27 15.36 14.30

S5

Table S4: Chemical composition of hemicelluloses derived from different biomass sources.

Hemicellulose

/composition Xylan

Extracted hemicellulose

from

corn stover

Minimally

damaged hemicellulose

of corn stalk

Mw, Da 10, 000 37, 345

PDI 1.03 1.58

Arabinose 0.5 5.95 8.3

Fucose — —

Galactose 1.3 7.5

4-O-methyl-glucuronic acid — 13.1

Galcuronic acid — —

Glucose 1.6 0.32 5

Mannose — 5.1

Rhamnose 0.8 —

Xylose 95.9 91.96 61

Ferulic acid — —

Acetyl — 6.2

Others 1.77

Ref. 2 1 3

S6

Table S5. Comparison of model yields (wt%) of major products from fast pyrolysis of native

hemicellulose xylan with different structures.

Hemicellulose structures

1,2-AXP

1,2;3,4-DAXP

Methyl glyoxal

GA Char CO2 H2O Acetol Acetic acid

V3 4.68 7.64 2.70 5.69 15.15 10.13 12.03 1.97 15.33

V4 3.12 8.44 2.75 7.00 14.37 9.61 11.94 1.57 15.20

V5 4.55 12.38 2.44 4.97 13.73 9.19 11.15 3.38 15.42

V6 4.68 4.37 2.69 5.66 16.37 10.94 12.98 0.56 15.33

Note that these are pure model predictions.

AXP=anhydroxylopyranose; DAXP=dianhydroxylopyranose; GA=glycolaldehyde.

V3: repeating units being xylose-acetylxylose, 31 xylose, 4 arabinose, 5 UA (4-O-methyl-

glucuronic acid), and 11 acetyl groups (attached to both position 3 and position 2).

V4: repeating units being xylose-acetylxylose, 31 xylose, 4 arabinose, 5 UA (4-O-methyl-

glucuronic acid), and 11 acetyl groups (attached to both position 3 and position 2).

V5: repeating units being xylose-acetylxylose, 31 xylose, 4 arabinose, 5 UA (4-O-methyl-

glucuronic acid), and 11 acetyl groups (attached to position 3 only).

V6: repeating units being xylose-acetylxylose, 31 xylose, 4 arabinose, 5 UA (4-O-methyl-

glucuronic acid), and 11 acetyl groups (attached to position 2 only).

S7

Scheme S1. Proposed structures of hemicellulose extracted from cornstover with varying

initial DP (degree of polymerization).

S8

Scheme S2: Proposed structures of extracted hemicellulose for decomposition into

monomeric, dimeric, or trimeric intermediates.

S9

Scheme S3. Decomposition mechanisms of native hemicellulose chains involving end-groups.

S10

Scheme S4. Decomposition mechanisms of native hemicellulose chains involving mid-groups.

S11

Scheme S5. Decomposition mechanisms of trimeric and dimeric species.

S12

Scheme S5 (continued). Decomposition mechanisms of trimeric and dimeric species.

S13

Scheme S5 (continued). Decomposition mechanisms of trimeric and dimeric species.

S14

Scheme S5 (continued). Decomposition mechanisms of trimeric and dimeric species.

S15

Scheme S6. Comparison of decomposition mechanism of 2-acetyl-xylopyranose and 3-acetyl-

xylopyranose.

S16

Scheme S7. Decomposition mechanisms of low molecular weight species.

S17

Scheme S8. Char formation of various dehydrated species.

S18

Scheme S9. Modified structures of native hemicellulose with varying amount of acetyl groups.

V1: 7 acetyl to 30 xylose; V2: 14 acetyl to 30 xylose; and V3: 11 acetyl to 31 xylose.

S19

Scheme S10. Modified structures of native hemicellulose with different repeating units. V3:

repeating units being xylose-acetylxylose, and V4: repeating units being xylose-acetylxylose.

S20

Scheme S11. Modified structures of native hemicellulose with acetyl group attached to different

positions of xylose units. V3: acetyl attached to both position 3 and position 2, V5: acetyl

attached to position 3 only, and V6: acetyl attached to position 2 only.

S21

Figure S1. Molecular structures and atom type and numbering scheme of β-D-glucopyranose

and β-D-xylopyranose.

S22

Experiments of fast pyrolysis of extracted hemicellulose by Zhang et al.1

(Adapted with permission from Jing Zhang, Yong S. Choi, Chang G. Yoo, Tae H. Kim, Robert C.

Brown, and Brent H. Shanks, Cellulose–Hemicellulose and Cellulose–Lignin Interactions during

Fast Pyrolysis, ACS Sustainable Chem. Eng., 2015, 3, 293–301. Copyright 2015 American

Chemical Society)

Materials and Methods. Hemicellulose was isolated from cornstover using the method

described below. Hemicellulose was extracted from cornstover by an aqueous ammonia treatment

followed by a hot water treatment.4 Cornstover, which was procured from the Agronomy Farm at

Iowa State University, was ground and screened to a nominal size of 9−35 mesh. Mineral

impurities are known to influence the product distribution during fast pyrolysis of the biomass.

The sieved cornstover was first acid washed to remove inorganic salts. Prior to using this

cornstover sample for pyrolysis studies, it was purified by washing with dilute acid. The cornstover

sample (≈10 g) was washed with 0.1 N HCl (100 mL) for 15 min, followed by two subsequent

washes with deionized water. The purified cornstover sample, containing <0.1 wt % ash, was used

throughout this study. The ash content of biomass samples was determined by subjecting the

sample (1 g) to combustion in a furnace at 575 °C for approximately 6 h. The carbon, hydrogen,

and nitrogen contents of the sample were analyzed by using a TruSpec CHN (Leco) instrument.

The oxygen content was determined by difference. Induced-coupled plasma mass spectrometry

(ICP-MS) was used to analyze the metal ion concentration in the cornstover sample.

Approximately 100 mg of cornstover sample was added into a 25 ml digestion tube with 10 ml of

concentrated nitric acid. The digestion tube was sealed and placed in a microwave for a pressurized

heating program: 1) 40 psi for 6 min; 2) 85 psi for 6 min; 3) 140 psi for 10 min. After digestion, 1

g of solution was taken out and diluted into 30 ml of 1 wt% nitric acid. The metal ions were then

S23

identified and quantified by ICP-MS after the equipment had been calibrated using standard

solutions. The metal ion content in cornstover was calculated by incorporating the dilution factor.

After acid washing to remove the mineral impurities, the cornstover was exposed to a 15 wt %

aqueous ammonia solution in a flow-through column reactor pressurized to 2.3 MPa. The reactor

was placed overnight in an oven set to a temperature of 170 °C to selectively cleave the ether bonds

in lignin for delignification. Exhaustive washing with DI water was performed after the aqueous

ammonia treatment. Then, the treated cornstover was firmly packed into a flow-through column

reactor, which had temperature and pressure control. A 0.07 wt % sulfuric acid aqueous solution

was passed through the reactor with flow rate of 5 mL min-1 at 180 °C under a pressure of 2.5

MPa. During the hot water treatment, the hydronium cation could initiate hemicellulose

depolymerization and cleave acetyl groups with the latter acting as a catalyst for further

depolymerization of hemicellulose. The depolymerized hemicellulose would enter the aqueous

phase thus being separated from the treated cornstover. The passed through solution was collected

and dried in vacuum at 50 °C to obtain solid particles. The solid was then acid washed with the

same condition as previously and ground into a fine powder.

Sample Characterization. Carbohydrate and lignin content in the biomass samples was

analyzed following the protocol from the NREL Chemical Analysis and Testing Stardard

Procedures: NREL LAP, TP-510-42618. Before quantification, biomass samples underwent two-

stage acid hydrolysis: (1) 72 wt % sulfuric acid for 1 h at 30 °C and (2) 4 wt % sulfuric acid for 1

h inside of an autoclave with the temperature held at 120 °C. Solid residues after the two-stage

hydrolysis were deemed acid insoluble lignin (AIL). Saccharides, which were in the liquid phase

after hydrolysis, were quantified using a HPLC with a Bio-Rad Aminex HPX-87P column (Bio-

Rad Laboratories, Hercules, CA) equipped with a refractive index detector. The acid soluble lignin

S24

(ASL) was quantified by measuring its absorbance at 320 nm in a UV−visible spectrophotometer.

Ash content in biomass was determined by oxidizing sample at 575 °C for 6 h inside of a

thermogravimetric analyzer (Mettler-Toledo Analytical).

Pyrolyzer-GC-MS/FID Experiments. The pyrolysis experiments were performed in a single-

shot micopyrolyzer (Model 2020 iS, Frontier Laboratories, Japan). Before pyrolysis,

approximately 500 μg of biomass was added to a deactivated stainless steel sample cup. The loaded

sample cup was then dropped gravitationally into a quartz pyrolysis tube. The pyrolysis

temperature of 500 °C was maintained by a tubular furnace surrounding the quartz reaction tube.

During the experiment, the generated volatile products were swept by the helium gas into a Bruker

430-GC through a deactivated needle. A capillary GC column, ZB-1701 (Phenomenex) was used

for separation of the volatile products. The column was either connected to a mass spectrometer

(MS, Saturn 2000) for product identification or to a flame ionization detector (FID) for product

quantification.

Product Identification and Quantification: The quartz pyrolysis tube inside of the single-shot

micopyrolyzer had an inner diameter of 4.7 mm and a length of 114 mm. Helium gas, which served

as the pyrolysis atmosphere and GC carrier gas, flowed through the pyrolysis system at a rate of

9.9 cm s-1. Before pyrolysis, the loaded cup was purged with helium for 30 seconds to remove

oxygen inside the cup then dropped into the heated furnace. As demonstrated previously,

maintaining a sample weight between 200-800 μg and particle size less than 75 μm ensured

negligible heat transfer and mass transfer limitation.5 The GC column contained a polar stationary

phase consisting of 14% cyanopropylphenyl and 86% dimethylpolysiloxane, with a length of 60

m, inner diameter of 0.250 mm and film thickness of 0.250 μm. The GC method consisted of an

injection temperature of 300 °C, split ratio of 100:1 and a constant carrier gas flow of 1 ml min-1

S25

through the column. The GC ramping program started with a 3 min hold at 35 °C, then increased

to 300 °C at a rate of 5 °C min-1 and finally was held at 300 °C for 4 min. To identify products,

the GC was connected to a mass spectrometer (MS) (Saturn 2000), using the electron ionization

mode with a 10 μamp emission current in the m/z ranging from 10 to 300. The mass spectra peaks

were compared with standard spectra of chemical compounds within the NIST library database.

Then the chemical identities were verified by running standards of the matching chemicals in the

same GC-MS system. As the column used had a similar stationary phase composition to previous

reports, the elution order of the chemical compounds in chromatogram was expected to match.6, 7

All the pure compounds used for peak confirmation were purchased from Sigma–Aldrich (US)

except 4-vinyl phenol, which was purchased from Alfa Aesar (US).

After product identification, a flame ionization detector (FID) was substituted for the MS for

product quantification. The FID was held at 300 °C with an air flow rate of 300 ml min-1 and

hydrogen flow rate of 30 ml min-1. The pure compounds used in product identification were also

used in calibration of the FID results. Standard solutions of the identified chemicals (except for

those discussed subsequently) were prepared by dissolving them in acetone, which eliminated

solute-solvent interaction for the chemicals.5 The levoglucosan calibration was performed by

pyrolyzing a known amount of levoglucosan in the micopyrolyzer-GC-FID system, which gave a

single peak corresponding to levoglucosan as verified by MS. For the glycolaldehyde calibration,

glycolaldehyde dimer was pyrolyzed at about 300 °C resulting in a single sharp peak in its

chromatogram, which was proven to be glycolaldehyde by MS. For dianhydro-xylose (DAXP 1,

DAXP 2, other DAXP 1 and other DAXP 2), which have a molecular weight of 114, a calibration

curve of a similar compound - 4-hydroxy-5-methyl-3-furanone (molecular weight 114) - was used

to determine their yields. For anhydro-xylanpyranose (AXP, other AXP), dianhydro-

S26

glucopyranose and levoglucosan-furanose, the calibration curve of levoglucosan was applied as

the proxy to estimate their yields.8 Eight-point straight line calibration curves (with R2≥0.95)

obtained by running duplicate standards at four different concentration levels (or sample weight)

were established for the pure compounds relating the peak area in the GC-FID chromatogram to

their respective standard concentration (or sample weight). As a result, forty eight compounds

(excluding char and gaseous compounds) were identified based on their mass spectrum (see Table

S1) and quantified.

To measure the CO and CO2 generated during pyrolysis, a De-Jaye gas analyzer equipped with

an infrared detector was connected to the split line of GC. The CO and CO2 concentrations were

recorded every second so the yield of CO and CO2 could be calculated by summing the amount of

gas generated over time using the known overall gas flow rate. The char yield was obtained by

weighing the sample cup before and after pyrolysis using a Mettler Toledo microbalance with a

sensitivity of 1 μg.

S27

Structure of the modeling codes

Figure S2. Flowchart of the modeling codes

The modeling codes were developed using programming languages Perl and C++ to

automatically generate the differential and associated algebraic equations, which were solved

using DDASAC numerical algorithms. Briefly, as shown in Figure S2, the first step is to specify

the reaction system in terms of reactor type, reactor volume, reaction conditions, and the properties

of the starting material, extracted hemicellulose. The structure of hemicellulose and derived chain

species can be expressed as a combination of a left-end group connected with a right-end group

S28

and/or a midgroup (Figure 2). Then the detailed reaction network of the decomposition of

hemicellulose and derived chain species and the reactions of low molecular weight (LMW) species

were specified. The rate equations and mass balance equations were generated using programming

languages Perl. The approach of continuous distribution kinetics was adopted here to describe and

simplify the decomposition dynamics of polymer chain species. The concentration of polymer

chains is modeled as a function of chain length and pyrolysis time, resulting in integral-differential

rate equations. Moment operations were employed to simplify and transform them into ordinary

differential equations (ODEs), implemented by the “Make moments” module. Then the “Make

inputs” module specifies the initials values of all the variables and parameters and collects all the

differential and associated algebraic equations for chain species and LMW species. Then the

“Make model” module translates all these into a “model.c” in C++ using Perl. Lastly, DDASAC,

a commercial software that solves nonlinear initial-value problems involving stiff implicit systems

of ordinary differential and algebraic equations, was used to solve the model and to provide species

profiles and other information.

S29

References

1. J. Zhang, Y. Choi, C. Yoo, T. H. Kim, R. Brown and B. H. Shanks, ACS Sustainable

Chemistry & Engineering, 2015, 3, 293–301.

2. J. Gu and J. M. Catchmark, Cellulose, 2013, 20, 1613-1627.

3. S. R. Wang, B. Ru, G. X. Dai, W. X. Sun, K. Z. Qiu and J. S. Zhou, Bioresour. Technol.,

2015, 190, 211-218.

4. T. H. Kim and Y. Y. Lee, Bioresour. Technol., 2006, 97, 224-232.

5. P. R. Patwardhan, J. A. Satrio, R. C. Brown and B. H. Shanks, J. Anal. Appl. Pyrolysis,

2009, 86, 323-330.

6. M. Garcia-Perez, S. Wang, J. Shen, M. Rhodes, W. J. Lee and C.-Z. Li, Energy Fuels, 2008,

22, 2022-2032.

7. G. Jiang, D. J. Nowakowski and A. V. Bridgwater, Energy Fuels, 2010, 24, 4470-4475.

8. P. R. Patwardhan, R. C. Brown and B. H. Shanks, Chemsuschem, 2011, 4, 636-643.

S30