Ring-locking enables selective anhydrosugar synthesis from carbohydrate pyrolysis †

The selective production of platform chemicals from thermal conversion of biomass-derived carbohydrates is challenging. As precursors to natural products and drug molecules, anhydrosugars are difficult to synthesize from simple carbohydrates in large quantities without side products, due to various competing pathways during pyrolysis. Here we demonstrate that the nonselective chemistry of carbohydrate pyrolysis is substantially improved by alkoxy or phenoxy substitution at the anomeric carbon of glucose prior to thermal treatment. Through this ring-locking step, we found that the selectivity to 1,6-anhydroβ-D-glucopyranose (levoglucosan, LGA) increased from 2% to greater than 90% after fast pyrolysis of the resulting sugar at 600 °C. DFT analysis indicated that LGA formation becomes the dominant reaction pathway when the substituent group inhibits the pyranose ring from opening and fragmenting into nonanhydrosugar products. LGA forms selectively when the activation barrier for ring-opening is significantly increased over that for 1,6-elimination, with both barriers affected by the substituent type and anomeric position. These findings introduce the ring-locking concept to sugar pyrolysis chemistry and suggest a chemical-thermal treatment approach for upgrading simple and complex carbohydrates.


Introduction
2][3][4][5] Pyrolysis is a widely studied thermal process for carbohydrate conversion, 1 but its usefulness in synthesis chemistry is limited by the low structural stability and high reactivity of carbohydrates. 6Anhydrosugars are primarily prepared from carbohydrate pyrolysis.Their fixed conformation, protected stereoselectivity, and retention of most of their chemical functionality make these sugar derivatives highly desirable precursors.][9][10][11][12] It also has strong potential as a sugar-based biorefinery feedstock chemical, with the US Department of Energy identifying it as a top-15 candidate. 13ound in smoke particles from forest fires, 14 LGA is currently prepared in small quantities by extensive purification from cellulose-derived pyrolysis oil. 1,11,15Large-scale production of LGA remains elusive.
To address this, we considered the counter-intuitive approach of pyrolyzing glucose.As the constituent sugar monomer of cellulose, glucose is precursor to a growing number of valuable chemicals, such as 5-hydroxymethylfurfural, 16,17 lactic acid, 18 butanol, 19 and γ-valerolactone (GVL). 5However, glucose pyrolysis generates mostly cracked products and char, with negligible amounts of LGA. 20In principle, synthesizing pure LGA from glucose pyrolysis can be accomplished by (i) carrying out the dehydration step that leads to intramolecular C 1 -O 6 bond formation, (ii) de-selecting the other dehydration pathways, and (iii) avoiding ringopening of the monosaccharide.
The length of cellulose plays a telling role in LGA formation in pyrolysis: longer chains generated more LGA than shorter ones. 20,21Mettler et al. carefully showed that cellulose with ∼130 sugar units yielded 27% LGA after pyrolysis, cellodextrins with 2-6 units yielded ∼8-18%, and D-glucose yielded 3%. 20onger chains contain a larger proportion of sugar units bound to another unit via the β-1,4-glycosidic linkage, from which we conjectured that a substituent group at the C 1 position improves LGA selectivity.In this work, we report that the pyrolysis of β-D-glucose ("β-G") generates LGA with excellent selectivity after its chemical modification at the C 1 position (Scheme 1).Using a combination of flash pyrolysis experiments and density functional theory (DFT) calculations, we show how LGA forms preferentially when ring-opening of the sugar unit is hindered by the presence of the C 1 substituent group and describe how the chemical environment around the C 1 atom impacts 1,6-anhydro bridge formation.

Thin-film for pyrolysis experiments
β-D-Glucose, methyl-β-D-glucoside, and methyl-α-D-glucoside were dissolved into deionized water (18.2MΩ, Barnstead Nano-pure Diamond System), while phenyl-β-D-glucoside and phenyl-α-D-glucoside were dissolved into 1-propanol : H 2 O (1 : 3 by volume) solvent mixture.4 μL of the solution (8 mg mL −1 ) was transferred into an open-end cylinder quartz tube (1.9 × 25 mm, diameter × length) by microliter syringe (Hamilton 700 series, 10 μL).Solvent was removed in a vacuum oven at 0.7 atm at 37 °C for 3 h, leading to the formation of a thin-film (containing 32 μg of sugar).The thin-film thickness is deter-mined from Scanning Electron Microscopy (SEM) which is equipped with Energy Dispersive X-ray Spectrometer (EDS) (see ESI for details †).SEM imaging with EDS indicated that the film thickness for all of the samples was less than 10 μm (Fig. S1 †).
To confirm appropriate reaction conditions, we followed the analysis of Mettler et al. 22 using the Pyrolysis number (Py) and Biot number (Bi) to evaluate our thin-film pyrolysis.Our calculations showed that the thin-film experimentation satisfied the criteria for kinetically limited, isothermal pyrolysis reactions (see the ESI †).

Pyroprobe coupled with GC/FID system
Thin-films pyrolysis experiments were conducted using a Model 5150 pyroprobe analytical pyrolyzer (CDS Analytical Inc.).The pyrolyzer was coupled to an Agilent 6890N GC/FID through a transfer line for online sampling (Fig. S2, see the ESI † for details of reaction and analysis system).2][23] The GC column was calibrated by injecting standard solution of methanol and phenol into the GC/FID system.LGA was calibrated by evaporating the LGA thin film sample directly through the pyrolyzer upon heating at pyrolysis temperature (600 °C), which appears as a single peak in the GC chromatogram. 21

Pyrolysis data evaluation
In order to monitor product distribution of glucose pyrolysis, the yields in this paper are reported in terms of molar carbon yield where the moles of carbon in the product are divided by the moles of carbon in the glucose moiety.
The following equations were used to calculate carbon yield of LGA (Y LGA ), conversion of different substrates (X sugar ) and selectivity to LGA (S LGA ).
LGA carbon yield (Y LGA ) for substituted glucose: LGA carbon yield (Y LGA ) from glucose was calculated by the same equation.
We used methanol and phenol as in situ generated internal standards to quantify glucoside conversion.Conversion (X MG ) for methyl-β-glucoside and methyl-α-glucoside were defined as follows: Conversion (X PG ) for phenyl-β-glucoside and phenylα-glucoside: For eqn ( 2) and ( 3), we confirmed that (i) methanol/phenol was thermally stable under pyrolysis condition (no side reaction); (ii) methanol/phenol did not react with other pyrolysis products to form ether; and (iii) pyrolysis products did not decompose to form methanol/phenol.
Conversion (X G ) for glucose was quantified using HPLC (details in next section): The LGA yield, conversion of starting precursor, and selectivity to LGA were related in the following manner: If i = MG, selectivity to LGA (S LGA ) is: If i = PG, selectivity to LGA (S LGA ) is: If i = G, selectivity to LGA (S LGA ) is:

Methyl-D-glucoside synthesis using glucose
The synthesis was carried out in a microwave reactor (Anton Paar Monowave 300), adapted from published procedures. 240 mg of glucose, 21.5 mg of Amberlyst 15 (dry) and 2 mL of methanol were transferred into a 30 mL reaction tube, which was then heated to 130 °C, and held for 30 min under stirring at 1200 rpm.After cooling, the solution was diluted with DI water to a total volume 10 mL, and then the solid resin acid catalyst was removed using a 0.2 mm microfiber syringe filter (25 mm, VWR).The filtrate containing crude methyl-glucoside was analyzed for methyl-α-D-glucoside ("α-MG") and methyl-β-D-glucoside ("β-MG") content by ion-exclusion high-performance liquid chromatography (HPLC), using a Shimadzu Prominence SIL 20 system (Shimadzu Scientific Instruments, Inc., Columbia, MD, USA) equipped with an HPX-87H organic acid column (Bio-Rad, Hercules, CA, USA) and a refractive index detector (RID).
The HPLC operation conditions were as follow: the column was operated at 42 °C, and 30 mM H 2 SO 4 mobile phase was flowing through the column at 0.
Four μL of filtrate was transferred into a quartz tube and dried under vacuum to form the crude methyl-glucoside thinfilm for the pyrolysis studies.
Glucose conversion in the synthesis step (X G_synthesis ): Y β-MG was calculated by following similar equations.
Methyl-D-glucoside has also been synthesized from cellulose, filter paper and cotton (see ESI † for details).

Computation method
Quantum mechanical and statistical mechanical computations were performed using the Gaussian 09 program package (rev C.01). 25 The classical B3LYP functional was used with the electronic wave function expanded by the 6-31G(d,p) basis set.The Gibbs free energy was evaluated at 600 °C based on frequency analysis at the same level of theory.This level of theory is at a relatively lower level compared to some previous studies.To compare with published results, the Gibbs activation energy (the free energy difference between the highest-energy transition state and the lowest-energy reactant 26 ) was computed to be 47.0 kcal mol −1 for LGA formation reaction, which was reasonably close to the value of 48.2 kcal mol −1 calculated from M06-2X/6-311+G (2df,p) level of theory 27 and the value of 52.5 kcal mol −1 from MP4(SDTD) single-point energy calculation based on geometries from B3LYP functional. 28

Results and discussion
As an initial test, we synthesized a methyl substituted glucose (MG) by replacing the C 1 hydroxyl group of glucose (β-G and α-G mixture) with a methoxy group through Fischer glycosylation chemistry. 29The resulting product contained 84 wt% methyl-D-glucoside, of which 44% was the β anomeric form.We heated the unpurified MG to 600 °C rapidly (flash pyrolysis ramp rate of ∼20 000 °C s −1 ), held it at 600 °C for 20 s, and analyzed the gas-phase effluent for pyrolysis products LGA selectivity from crude MG pyrolysis was ∼64%, significantly higher than the experimentally determined ∼2% selectivity from glucose pyrolysis.We next carried out the pyrolysis of high-purity methyl-β-Dglucoside ("β-MG") and methyl-α-D-glucoside ("α-MG") under the same experimental conditions.We chose 600 °C for our studies, which is the higher end of the temperature range used in fast pyrolysis processes. 30At lower pyrolysis temperatures, the selectivity for anhydrosugar increases and the selectivity for light oxygenates and permanent gases (CO 2 and CO) decreases (Table S1 †).We adopted the thin-film pyrolysis technique of Dauenhauer and co-workers 22 to ensure the various sugar compounds did not experience concentration and temperature gradients during heating and to ensure consistent substrate conversion and LGA yield measurements.Fig. 1 shows representative chromatograms of the detected products from the pyrolysis of glucose, β-MG, and α-MG, as well as the β and α forms of phenyl-D-glucoside (PG).2][23] In comparison, pyrolysis of the substituted sugars led to mostly LGA (with methanol or phenol eliminated as a co-product) with decreased amounts of the other pyrolysis products.
A direct comparison of LGA yields provided an incomplete analysis, because the sugar compounds volatilized to different extents during pyrolysis treatment and did not undergo complete conversion (<60%).Instead, we quantified and compared LGA selectivity values, which accounts for incomplete conversion of the parent sugar (Table 1).Fig. 2 shows LGA selectivities from MG and PG pyrolysis, which readily exceeded 20%.In comparison, glucose pyrolysis had very low LGA selectivity (∼1-4%) at all conversions observed (∼30-100%).LGA selectivity from β-MG pyrolysis was in the 80-85% range at <20% con-  version and it decreased to ∼48% near 50% conversion.Furan derivatives, light oxygenates, and char residue formed at higher conversions.α-MG pyrolysis showed a similar trend but selectivities were roughly half as much.Pyrolysis of the phenoxy-functionalized glucoses gave similar trends, with LGA selectivities measured as high as 93 ± 4% for β-PG.The highest LGA yield (39%) came from β-PG pyrolysis at 54% conversion (Table 1).At a sugar/glucoside conversion of 35%, LGA selectivity decreased in the following order: β-PG > β-MG > α-PG > α-MG ≫ glucose, showing the wide-ranging effects of substituent type and anomeric position.We next studied the energy profiles of several key reaction pathways in β-G, β-MG, and β-PG pyrolysis, through DFT computation using the B3LYP functional and 6-31G(d,p) basis set.β-G can dehydrate into LGA and other anhydrosugars. 27,33It can also mutarotate to the open-chain aldehyde form, which then goes through isomerization, retro-aldol cleavage, and dehydration reactions.Consistent with observations of numerous cracked products and low LGA selectivity, the ring-opening pathway for β-G pyrolysis has the lowest activation energy (P-1; 37.3 kcal mol −1 ), followed by 1,6-dehydration to form LGA (P-2; 47.0 kcal mol −1 ) and then by dehydration via β-eliminations (P-3 to P-5; ∼59-71 kcal mol −1 ) (Fig. 3).
Of particular interest are the lowest-barrier P-1 and P-2 pathways, which proceed through a common series of conformational changes of the pyranose ring.β-G flips from its stable chair conformation I into a boat structure III and then into an inverted chair V; the energy barriers are sufficiently small (∼8 kcal mol −1 ) such that these structures can be considered in pre-equilibrium with one another (Fig. 4a, Table S3 †).Through a concerted set of bond cleavage and formation steps, structure V then converts into either the openchain form via a six-atom-containing transition state VI 34 or into LGA via a four-atom-containing transition state vi. 35ith the C 1 methoxy substituent, β-MG goes through P-1 and P-2 pathways via geometries very similar to those of β-G (Fig. 4b, Table S3 †).The relative energies of the various states are essentially the same as in the β-G case, except for transition state VI.The activation barrier height increases substantially (by ∼50 kcal mol −1 to 86.2 kcal mol −1 ), leading to a ringlocking effect.Ring-opening becomes very difficult because the C 1 methoxy impedes the formation of the transitory sixatom cyclic structure.Whereas the 1s orbital of the H atom on O 1 of β-G has ample overlap with the O 6 sp 3 orbital (Fig. 6a), the methyl carbon sp 3 orbital on O 1 of β-MG does not (Fig. 6b).The methoxy group lowers the P-2 and P-3 barriers somewhat, and has no effect on the much higher P-4 and P-5 barriers.The observed high LGA selectivities directly result when 1,6-elimination becomes the energetically favored pathway after ring-opening is greatly suppressed.
We found that high LGA selectivities come primarily from the ring-locking effect by the C 1 group, even with the α anomers.Ring-opening of α-G, α-MG and α-PG proceeds through a different pathway involving a four-atom cyclic structure (Fig. S4 and S5 †). 34The activation energy increases from 44.9 kcal mol −1 to 76.2 and 70.1 kcal mol −1 when α-G is functionalized at the C 1 with methoxy and phenoxy groups, respectively.
LGA formation is more difficult when the C 1 OH/OR groups and C 6 hydroxymethyl groups are on opposite sides of the pyranose ring, with the α anomers going through a two-step carbene mechanism instead of the four-atom cyclic structure of the P-2 pathway. 27The C 1 substituent abstracts the hydro-  Mannosan selectivities were 5% and 50%, respectively, after accounting for the difference in conversion.Similarly, pyrolysis of D-galactose and methyl-α-D-galactoside led to 5% and 11% galactosan yields and to selectivities of 5% and 23%.
Ring-locking can explain the higher LGA yields from longer cellulose chains than shorter ones.Cellulose pyrolysis generates LGA through depolymerization, in which the glycosidic bond between the C 1 of one sugar unit and the O 4 of an adjacent unit cleaves, and C 1 forms the intramolecular bond with O 6 .The adjacent unit at the C 1 position helps hinder ring-opening during depolymerization.It can also help explain how lignocellulose pyrolysis leads to higher LGA selectivities normalized to cellulose content (∼48% for poplar wood v. ∼27% for Avicel cellulose 37 ).The cellulose fraction contains longer chains and has additional glycosidic linkages to hemicellulose and lignin, both of which contribute to the ringlocking effect.We studied if cellulose solids can be pre-treated in such a way to improve LGA selectivity prior to pyrolysis.As a proof-ofconcept, three different cellulose sources (microcrystalline cellulose, filter paper and cotton) were methoxylated into mixtures of α-MG and β-MG (Table S4 †).Flash pyrolysis of the cellulosederived methyl-D-glucoside crude gave a LGA yield of 15%.Pyrolysis of untreated cellulose gave a similar yield, though needing complete cellulose conversion.Accounting for the differences in conversion, LGA selectivities were 59% and 14% for treated and untreated cellulose, respectively.LGA selectivities (∼50-60%) from the various methyl-D-glucoside crudes mapped between the selectivity-conversion curves for α-MG and for β-MG (Fig. S7 †).Chemical modification of the carbohydrate source reduces the atom-inefficiency of cellulose pyrolysis.

Conclusions
Ring-locking is established as an efficient means to achieve substantially improved LGA selectivities from the pyrolysis of sugars.The presence of a substituent group at the anomeric carbon impedes pyranose ring-opening by raising the activation barrier of this pathway.The barrier for 1,6-elimination to form LGA decreases when the substituent group has a strong inductive effect.The multitude of reaction pathways in pyrolysis may be selectively redirected for substantially improved selectivities to anhydrosugars by simply ring-locking the anomeric carbon of the carbohydrates.Similar methods may also be applicable to monosaccharides with furanose ring structures.

Fig. 2
Fig.2LGA carbon selectivity-conversion analysis for the different sugar compounds.Each data point represents a minimum of 3 experiments, and the error bar is ±1 standard deviation.LGA selectivity values (S LGA = Y LGA ÷ X sugar ) were determined from yields Y LGA measured at different sugar conversions (X sugar ).Conversion varied with carrier gas contact time at 600 °C.
gen at C 1 to form a three-center transition state (structure b), resulting in a carbene intermediate (structure c); the latter then forms LGA via another three-center transition state (structure d).The LGA formation activation energy is relatively large, changing little after methoxy functionalization (from 66.2 to 69.0 kcal mol −1 ) and decreasing after phenoxy functionalization (to 51.4 kcal mol −1 ) (Fig. S4 and S5 †).The lower LGA selectivities for α-MG and α-PG compared to β-MG and β-PG are due to the smaller activation barrier differences between ring-opening and 1,6-dehydration/elimination.We observed the ring-locking effect in other C 1 -functionalized monosaccharides (Fig. S6 †).D-Mannose and D-galactose are common stereoisomers of D-glucose, in which their respective C 2 and C 4 OH groups are in the axial position.Pyrolysis of D-mannose (a mixture of α and β forms) led to 5% yield of the corresponding 1,6-anhydrosugar mannosan; in comparison, pyrolysis of methyl-α-D-mannoside led to 23% yield.

Fig. 4
Fig. 4 Relative Gibbs free energy profiles for ring-opening (P-1) and LGA formation (P-2) pathways for (a) β-G, (b) β-MG and (c) β-PG pyrolysis at 600 °C.Structures vi and VI are the rate-determining transition states containing transient four-atom and six-atom cyclic structures, respectively.3-D structures for (a) is shown in Fig. 5, for (b) and (c) are shown in Fig. S3.†

Fig. 5
Fig. 5 Representative 3-D structures used in β-G ring opening and LGA formation pathway calculation, optimized at the B3LYP/6-31G(d,p) level of theory.

Table 1
LGA conversion, yield, and selectivity values from glucose/glucoside pyrolysis at 600 °C.The bolded values were used in the pyrolysis experiments of Fig.1 α-PG (predicted boiling point = 482.1 ± 45.0 °C)