Functional group dependence of the acid catalyzed ring opening of biomass derived furan rings: an experimental and theoretical study

Abstract

We describe studies of Brønsted acid catalyzed ring opening of substituted furans contained within biomass derived C₈- and C₉-molecules. Ring opening occurs homogeneously under relatively mild conditions of 80 °C using catalytic hydrochloric acid. In the case of 4-(5-methyl-2-furyl)-2-butanone (1a), the reaction proceeds to a single product in up to 92% yield after 24 hours. For 4-(2-furanyl)-2-butanone (1b) and 4-(5-hydroxymethyl)-2-furanyl-2-butanone (1c), however, multiple products are observed, illustrating the significant influence of furan ring substituents on the reactivity of this class of compounds. The generality of these reaction pathways was tested using several other similar substrates. Kinetics experiments indicate that ring opening of 1a occurs via specific acid catalysis, and computations elucidate the effect of initial protonation on the reaction pathway. Calculated pK_a values were calibrated against experimentally measured values and are consistent with observed reactivities. Inclusion of explicit, hydrogen-bonded water molecules in addition to the SMD solvent model is necessary when studying protonation of alcohol and ketone groups.

---

Introduction

Conversion of non-food based plant biomass into transportation fuels and chemical feedstocks has recently received a great deal of attention.¹ To achieve this, several challenging chemical transformations must be performed, including the breakdown of lignocellulose, followed by (in the case of high energy density fuel synthesis) lengthening of carbon chains within the resulting sugars, and then extensive deoxygenation and hydrogenation of the resulting products.¹ Current technologies to convert biomass into fuels include gasification followed by Fischer–Tropsch synthesis and pyrolysis to bio-oil.^1–3 Unfortunately, these processes are generally non-selective and require high temperatures as well as initial oxidation of the biomass.

Cellulosic biomass can be hydrolyzed under mild conditions yielding five and six carbon sugars as products.⁴ Subsequent dehydration of these sugars affords products containing furan and aldehyde moieties, including furfural, methyl furfural, dimethyl furan and hydroxymethyl furfural (HMF).^4,5 An aldol condensation between these furfural products and acetone selectively yields compounds containing eight or nine carbon atoms⁶ that are suitable for upgrading to diesel fuels. These and similar compounds can be ring opened, deoxygenated and hydrogenated by a variety of heterogeneous metal catalysts at high temperatures and high hydrogen pressures to provide linear alkanes suitable for use as fuels.^1,3,7

The petrochemical industry has studied the conversion of crude oil into fuels and synthetic feedstocks for more than a century,⁸ resulting in an enormous body of knowledge and understanding of the processes involved. In contrast, the understanding of biomass conversion to hydrocarbons is comparatively limited. Furan rings, for example, are ubiquitous in cellulose derived fuel precursors, and opening these rings is necessary for production of the desired linear alkanes.¹ Many of the conversion processes to date view the furan ring simply as an aromatic to be hydrogenated. Ring opening the resulting tetrahydrofuran then requires comparatively extreme conditions, as illustrated in Path A of Scheme 1,^1,3,7 or stoichiometric coupling with other reagents.⁹ Species 1 in this case represents the result of a commonly adopted approach in which the furan ring and the ketone are both hydrogenated. High temperature, heterogeneous processes have been shown to be effective for furan ring opening reactions; however wide product distributions and fragmentation of the carbon backbone is often observed.^7,10 Development of selective processes that occur at lower temperature and pressures will reduce capital and process costs, making the conversion of biomass into fuels more economically viable. Furan ring opening reactions using metal catalysts and stoichiometric oxidants have also been studied in detail,¹¹ as has Pt-catalyzed electrolytic furan ring opening,¹² but these approaches are impractical for fuel production.

Scheme 1

In contrast, the direct ring opening of the furan ring is a good first step since the acid catalyzed ring opening occurs readily under relatively mild conditions (Path B, Scheme 1). Acid catalyzed furan ring opening has been known for some time,^13,14 but use of this pathway for hydrocarbon production has not been considered to any great extent and has received relatively little detailed mechanistic study, especially with respect to biomass derived substrates. This enables a lower temperature, chemically selective approach that utilizes inexpensive catalysts for the upgrading of biomass, which is very desirable. Following acid catalysed furan ring opening, the biomass could be treated using the same hydrodeoxygenation processes required in Path A. Furthermore, development of a detailed understanding of the mechanism and selectivity of furan ring opening in biomass derived substrates will enable design of improved processes and better catalysts. We will report on more recent results pertaining to the conversion of species such as 2 below into hydrocarbons in due course.¹⁵

Herein, Brønsted acid catalyzed ring opening of biomass derived molecules containing eight and nine carbon atoms are presented. Ring opening occurs under mild conditions (80–100 °C) using catalytic HCl. In the case of 4-(5-methyl-2-furyl)-2-butanone (1a), the reaction proceeds quantitatively to a single product. For 4-(2-furanyl)-2-butanone (1b) only slight decomposition is observed while for 4-(5-hydroxymethyl)-2-furanyl-2-butanone (1c) multiple products are observed, indicating the significant effect of substituents on the reactivity of this class of compounds. The α,β unsaturated precursors 3a–c were also tested for ring opening, but only slight decomposition was observed. DFT calculations along with kinetics experiments indicate a specific acid catalysis mechanism and show that the location of initial protonation influences the overall product distribution.

Results

Ring opening studies began with the nine-carbon containing species 4-(5-methyl-2-furyl)-2-butanone (1a), shown in Scheme 2. 1a was obtained via aldol condensation between 5-methylfurfural (a product of cellulose hydrolysis and dehydration⁵) and acetone,^6,16 followed by reduction of the exocyclic unsaturation with magnesium.¹⁷ Heating 1a at 100 °C with excess HCl has been reported to give 2,5,8 nonanetrione (2a) in high yield,¹⁸ so this reaction served as a starting point for development of a catalytic process under milder conditions. Initial test reactions using a ten-fold excess of HCl at 100 °C over 3 hours did show clean conversion to 2a in 93% yield determined by ¹H NMR spectroscopy. Following these results, milder conditions with lower temperatures and catalytic amounts of HCl were investigated. 1a is only slightly soluble in water, so five different mixtures of water and water-miscible organic solvent (1 : 1 by volume) were also tested. Control reactions showed no reaction after heating 1a in the absence of catalyst at 80 °C for 48 hours. As illustrated in Table 1, good yields of 2a were observed in several cases after 24 hours, even after heating at only 60 °C, and no other products were observed. Optimal conversion to 2a was observed at higher temperatures and while all the organic solvent mixtures made 1a soluble in the reaction mixture, conversion to 2a was not greatly improved relative to pure water as the reaction solvent at 100 °C.¹⁹ Throughout the course of these reactions, no intermediates were observed; only starting material and the final product were detected by ¹H NMR spectroscopy and GC-MS. When this reaction was performed on a half gram scale, 2a was obtained in 85% isolated yield.

Scheme 2

Table 1 Yield of ring opened product 2a from heating 1a for 24 hours in various solvent systems with 10 mol% HCl

Solvent^a	60 °C	70 °C	80 °C
a Solvent mixtures are 1 : 1 water : organic solvent by volume.
Water : acetone	12%	25%	59%
Water : acetonitrile	18%	43%	77%
Water : methanol	29%	59%	94%
Water : ethanol	14%	25%	50%
Water : isopropanol	11%	22%	41%
Water	44%	91%	87%

---

Other Brønsted acids were also tested as catalysts for this reaction, as shown in Table 2. The strong Brønsted acids HNO₃, H₂SO₄ and CF₃COOH all give yields of 2a similar to those observed when HCl is used. In contrast, the weaker organic acids, acetic acid and formic acid, give only 1.4% and 4% yields of 2a, respectively after 24 hours. A variety of metal salts were also tested as catalysts, but none were as effective as the strong Brønsted acids discussed here.²⁰

Table 2 Yield of ring opened product 2a using 10 mol% catalyst in 1 : 1 water : methanol as solvent

Acid	Yield of 2a
HCl	94%
HNO3	77%
H2SO4	84%
CH3COOH	1.4%
HCOOH	4%
CF3COOH	77%

---

Several experiments were performed to further probe the mechanism of the ring opening reactions. Compound 1a was heated in solutions buffered to a pH of 3 with a 1 : 1 ratio of HCl and KCl, with buffer concentrations of 0.2 M, 0.16 M, and 0.12 M. All three solutions gave the same yield of ring opened 2a after heating at 80 °C for 24 hours, despite nearly halving the buffer concentration. This is consistent with a specific acid catalysis mechanism in which the rate of catalysis depends solely on the concentration of hydronium ions in solution. Compound 1a was also heated for 24 hours at 80 °C in a 1 : 1 mixture of CD₃OD and D₂O with 10 mol% of (proteo) hydrochloric acid as a catalyst. After 24 hours, the observed yield of 2a was only 45%, likely due to the lower acidity of D₂O solutions.²¹¹H and ¹³C NMR analysis of the product indicate that deuterium is incorporated predominantly in the methylene positions of the product, though some deuteration of the methyl groups is also observed.

A recent study has highlighted the effect of different furan ring substituents on observed reactivity,²² so the ring opening of other substrates was also examined here. The eight carbon compound 4-(2-furanyl)-2-butanone, 1b, was prepared using the same method as for 1a using furfural and acetone.^16,17 Approximately 5–10% decomposition to unknown products is observed upon heating 1b at 80 °C with 10 mol% of HCl in a 1 : 1 water : methanol mixture (Scheme 3). In contrast, heating 1b in a 1 : 1 water–acetone mixture yields a nineteen-carbon product 2b in 90% yield by ¹H NMR spectroscopy, as shown in Scheme 4. Running the reaction in D₂O and acetone-d₆ gives the same product, but with incorporation of ca. 10 deuterium atoms: based on ¹H NMR analysis, the methyl groups between the furan rings are completely deuterated while significant amounts of deuterium are present in the ketone methyl groups as well. Thus, 2b appears to be the product of condensation of two molecules of 1b and one acetone molecule (Scheme 4). No reaction is observed when 1b is heated in water : acetone solvent without added acid. Heating 1b in other solvent mixtures gave complex mixtures of unknown products.

Scheme 3

Scheme 4

A third biomass derived substrate 4-(5-hydroxymethyl)-2-furanyl-2-butanone, 1c, was prepared from HMF and acetone,¹⁶ followed by reduction of the exocyclic unsaturation with magnesium.¹⁷ Upon heating 1c in water or organic solvent mixtures, several products are observed by ¹H NMR spectroscopy and GC-MS. One major product (ca. 50%) is the partially deoxygenated α,β-unsaturated compound 3a (Scheme 5). Column chromatography yielded complex mixtures of other unidentified products. The absence of furan peaks in the ¹H NMR spectra of some of these fractions suggests that some ring opening does occur, however. No reaction is observed when 1c is heated in water : acetone solvent without added acid.

Scheme 5

Sodium borohydride reduction of 1a yielded the ketone reduced compound, 4-(5-methylfuran-2-yl)butan-2-ol (1d). Heating 1d in the presence of HCl gave the ring opened diketone alcohol 2d quantitatively as determined by ¹H NMR spectroscopy (Scheme 6). Sodium borohydride reduction of 1b gave 4-(furan-2-yl)butan-2-ol (1e), and slight decomposition was observed when 1e was heated with HCl (Scheme 7). Thus, reduction of the ketone to the alcohol does not appear to affect the reactivity with respect to furan ring opening. Ring opening of the α,β unsaturated products 3a–c was also investigated as summarized in Scheme 8. Upon heating compounds 3a–c in 1 : 1 water : methanol at 80 °C, some darkening of the reaction mixture occurred, but no ring opening is observed by NMR after 24 hours.²³ Additionally, no condensation is observed when 3b is heated in acetone, so the reaction shown in Scheme 4 is unique to the hydrogenated compound 1b.

Scheme 6

Scheme 7

Scheme 8

Computational results

The thermodynamics of the reactions discussed above were investigated via DFT calculations using the B3LYP²⁴ density functional and 6-311+G(d) basis set or the high-level CBS-QB3²⁵ composite method. The conductor-like polarizable continuum model (CPCM) was used to approximate the effects of solvation.²⁶ Natural atomic charges²⁷ were computed with the NBO 3.1 program as implemented in Gaussian 09.²⁸ The method of Liptak et al. was used, with a few modifications (vide infra) to calculate pK_a values in solution.²⁸ Although ring opening reactions were carried out at 80 °C experimentally, free energies were computed at 25 °C for two reasons: first, Liptak's method incorporates an empirical proton solvation free energy measured at 25 °C,²⁹ and second, free energies for overall ring opening (Table 3) increase by no more than 1.8 kcal mol⁻¹ at 80 °C relative to those calculated at 25 °C. All calculations were performed using the Gaussian 09 program.²⁷

Table 3 Thermodynamics of overall ring opening reaction, calculated using the CPCM solvation model at 298 K

Reaction	ΔG^o (kcal mol^−1)
	B3LYP/6-311+G(d)	CBS-QB3	CBS-QB3 (80 °C)
1a + H2O → 2a	−4.92	1.65	3.34
1b + H2O → 2b*	−2.03	3.45	5.11
1c + H2O → 2c	−4.44	0.28	2.03
1d + H2O → 2d	−4.88	0.77	2.36
1e + H2O → 2e	−2.88	2.90	4.48
3a + H2O → 4a	2.88	5.61	7.10
3b + H2O → 4b	4.34	7.24	8.76
3c + H2O → 4c	8.27	3.95	5.55

---

First, the overall thermodynamics of furan ring opening were examined. Structures of compounds 1a–e, 2a, 2b*, 2c–e, 3a–c and 4a–c were first optimized in the gas phase (where 2b*, 2c, and 2e are the ring opened products of furans 1b, 1c, and 1e, respectively²⁰) followed by re-optimization using the CPCM solvent model (Table 3). With B3LYP/6-311+G(d), ring opening of 1a–e to give the corresponding triketones is exergonic by 2.03 to 4.92 kcal mol⁻¹. In contrast, ring opening of the α,β unsaturated substrates 3a–c is endergonic by 2.88 to 8.27 kcal mol⁻¹. This is consistent with the experimentally observed reactivity: the hydrogenated substrates 1a–c ring open or decompose slightly under the experimental conditions whereas the α,β unsaturated substrates 3a–c do not. Calculations with the CBS-QB3 method give somewhat different results, however. Ring opening of 1a–e is slightly endergonic (0.28 to 2.90 kcal mol⁻¹) while ring opening of compounds 3a–c is even more unfavorable (3.95 to 7.24 kcal mol⁻¹). In all cases, the ring opening is more endergonic with CBS-QB3 relative to B3LYP, with the exception of 3c (8.27 (B3LYP) vs. 3.95 (CBS-QB3) kcal mol⁻¹): this anomaly arises from enhanced solvent stabilization of 3c with B3LYP relative to the other furans and ring opened products.²⁰ Despite the difference in absolute energies between the two methods, the qualitative result is the same: ring opening of the α,β unsaturated substrates is less favored thermodynamically.

Next, pK_a values were calculated for select carbon and oxygen atoms of compounds 1a–e and 3a–c. In Liptak's method,²⁸ pK_as are calculated with the CBS-QB3 composite method and CPCM solvent model using a thermodynamic cycle approach involving gas phase (CBS-QB3) acid dissociation (HA → A⁻ + H⁺) and solvation (CPCM). However, whereas Liptak's analysis is limited to dissociation of neutral phenols to their anionic conjugate bases (HA → A⁻ + H⁺), the pK_a values in the present work correspond to dissociation of protonated furans to their neutral conjugate bases (BH⁺ → B + H⁺). Due to this disparity, the solvent model and level of theory for solvation free energies were recalibrated against the experimental pK_a of 2,5-di-tert-butyl furan (Table 4).³⁰ In addition to the methods used for ring opening thermodynamics (CBS-QB3 and B3LYP/6-311+G(d) with CPCM solvation), Truhlar's universal SMD solvation model was employed along with three methods over which the model was optimized (B3LYP/6-31G(d), M05-2X/6-31G(d), and HF/6-31G(d)).³¹ The most accurate result relative to the experimental value was obtained for HF/6-31G(d), the simplest method in the calibration set, in conjunction with CPCM (pK_a = −10.6). This result is not particularly surprising, given that the CPCM solvent cavities were optimized for HF and small basis sets.^25,28 Furthermore, Liptak²⁸ found solvation free energies at this level of theory (CPCM/HF/6-31G(d)) to also yield the most accurate pK_as of substituted phenols, suggesting that the best method for CPCM solvation within Liptak's method may be insensitive to the nature of the chemistry (protonation of neutral species vs. dissociation of acids) and identity of reagents (furans vs. phenols).

Table 4 Calibration of computational model for solvation free energies within pK_a calculations as a difference (ΔpK_a) between calculated and experimentally observed pK_a of 2,5 di-tert-butylfuran. Values correspond to protonation at the tertiary carbon

Method	Solvent model	pKa	ΔpKa
Experimental value	—	−10.0	—
CBS-QB3	CPCM	−13.1	−3.1
CBS-QB3	SMD	−7.5	2.5
B3LYP/6-311+G(d)	CPCM	−11.1	−1.1
HF/6-31G(d)	SMD	−4.9	5.1
HF/6-31G(d)	CPCM	−10.6	−0.6
B3LYP/6-31G(d)	SMD	−5.8	4.2
B3LYP/6-31G(d)	CPCM	−11.3	−1.3
M05-2X/6-31G(d)	SMD	−6.2	3.8
M05-2X/6-31G(d)	CPCM	−11.5	−1.5

---

Use of a single experimental reference pK_a value is sufficient for this study for several reasons. First, the differences between calculated pK_a values of 1a–e and 3a–c using either CPCM/HF/6-31G(d) or CPCM/B3LYP/6-311+G(d) solvation energies are not drastic, with most values differing by no more than 4 pK_a units.²⁰ Additionally, the latter of the two above methods gave the second most accurate pK_a of 2,5-di-tert-butyl furan (Table 4). Finally, the most acidic site of each molecule is in the same position regardless of the method used.²⁰ This illustrates the difficulty of calculating accurate pK_a values, and more extensive experimental and theoretical study of protonation of furan derivatives would be of value.

In terms of solvent model, SMD gives less accurate pK_a values than CPCM, even though the former was found to yield more accurate solvation free energies.³⁰ Whether or not this finding is indicative of deficiencies in Truhlar's SMD method is unclear, however, particularly given the scarcity of experimental pK_a data for furans. Nevertheless, CPCM/HF/6-31G(d) solvation energies will be used for most of the pK_a calculations in this work.

The pK_a values for protonation of select atoms within furans 1a–e and 3a–c are shown in Scheme 9. For 1a, the most basic site of the furan ring is the carbon atom in the 2 position with a calculated pK_a of −13.0. This calculated value is close to the experimental pK_a value of −10.01 for 2,5-di-tert-butyl furan,²⁹ which has a very similar furan moiety. The remaining furan carbons are only slightly less basic with pK_a values ranging from −14.7 to −18.1. The furan oxygen is much less basic with a pK_a of −33.1.

Scheme 9 Calculated pK_a values for dissociation of a proton from the conjugate acids of substrates 1a–e and 3a–c (HA⁺ → A + H⁺). *Denotes proton migration during geometry optimization, preventing calculation of a pK_a value for that atom. ^aDenotes an explicit water molecule was added and SMD/HF/6-31G(d) solvation free energies are used in place of CPCM/HF/6-31G(d) during calculation of the pK_a value.

Initially, we calculated that the ketone oxygen of 1a has a pK_a value of −18.8. This was surprisingly different from the experimentally measured pK_a of protonated acetone, which has a pK_a between −2 and −3.³² Our computational method gave a pK_a of −17.6 for protonated acetone, almost 15–16 pK_a units less basic than the experimental value. This suggests a deficiency in the method, and so the effect of an explicit water molecule hydrogen bonded to the protonated carbonyl group was tested.¹⁹ For acetone, addition of just one explicit water molecule yields a calculated pK_a of 8.6, a value closer to the experimental value but significantly more basic than the measured value. Further improvement is found when SMD/HF/6-31G(d) solvation free energies are used in place of CPCM/HF/6-31G(d) values, giving a calculated pK_a of 1.8. This trend is also observed for 1a, where the calculated pK_a goes from −18.8 to 7.6 upon addition of a water molecule, and 7.6 to 0.4 on switching to SMD solvation. This behaviour is in contrast to furan ring protonation (Table 4), in which CPCM gives more accurate pK_as than SMD. Since this trend is also found for the protonated carbonyl and hydroxyl groups of the remaining substrates,¹⁹ pK_as reported for these groups will correspond to deprotonation with an explicitly hydrogen-bonded water (FuranH⁺ − H₂O → Furan + H₃O⁺) with SMD solvation, unless otherwise stated (Scheme 9). These considerations do not seem as important for the furan ring, however, where the calculated pK_a values appear to be in good agreement with experiment without explicit water molecules added. While use of different solvation models for different functional groups in the same substrate may seem problematic, it is appropriate in this case for two reasons. First, this dual methodology yields pK_as close to expected experimental values (see above) and second, even with CPCM solvation, the hydroxyl and carbonyl groups remain the most basic sites for each substrate.¹⁹ Replacing the electron-donating furyl methyl group of 1a with a proton gives 1b. The comparatively electron withdrawing proton makes the furan ring carbons within 1b less basic than those of 1a by 2–6 pK_a units with the exception of C5, which is the most basic site on 1b. Notably, protonation at C5 followed by geometry optimization yields a tertiary carbocation with the positive charge localized on the C2 carbon via rearrangement of the ring π-electrons, as shown in Scheme 10. This was confirmed by computing CPCM/B3LYP/6-311+G(d) natural atomic charges for both protonated and neutral 1b. While C5 protonation results in a decrease of positive charge on C5 relative to 1b (Δq = −0.196; Scheme 10), the positive charge increases on both C2 and O by 0.295 and 0.067 units, respectively. In contrast, initial protonation at the C2 carbon followed by geometry optimization gives a comparatively less stable secondary carbocation with the positive charge located on the C5 carbon (Scheme 10). Once again, the positive charge decreases at the site of protonation (C2) by −0.242 units and increases on C5 and O by 0.334 and 0.089 units, respectively (Scheme 10); this increase is greater than that on C2 and O upon C5 protonation, implying that C2 protonation gives a less stable carbocation. These differences in carbocation stability account for the relatively large difference in pK_a values calculated for C2 (−17.0) and C5 (−13.3) in 1b; the difference in pK_a value is much smaller in 1a where tertiary carbocations are obtained via protonation at C2 or C5. The pK_as of the furan and ketone oxygen atoms do not differ substantially from those in 1a. As seen in 1a, the calculated pK_a of the carbonyl oxygen in 1b increases from −18.9 to 0.1 upon addition of an explicit water molecule and use of SMD solvation.

Scheme 10 Rearrangement of π electrons upon protonation of 1b at the 2 and 5 position of the furan ring. Values of Δq, the difference in natural atomic charges between protonated and unprotonated 1b, are highlighted in blue.

The condensation reaction between 1b and acetone (Scheme 4) is consistent with the calculated pK_a values. Under the experimental conditions, some of the solvent acetone molecules will be protonated, and the relatively nucleophilic C5 atom within 1b can then attack the electrophilic central acetone carbon. While compounds 1a, 1c, and 1d have similarly basic C5 carbons, they may be too sterically crowded to undergo the same reaction. Interestingly, 1a yields the ring opened triketone on heating with HCl (Scheme 2) while 1b does not react under these conditions (Scheme 3). Whether this discrepancy in chemistry arises from differences in the position of initial protonation or the presence of a furan methyl group (or both) will be addressed in a forthcoming theoretical analysis of the reaction mechanism.

The terminal hydroxyl group in 1c significantly influences the pK_a values within the furan ring. With the exception of C5 and the ketone oxygen, calculated pK_a values of the furan carbon atoms are ca. 1 to 3 pK_a units less basic than in 1a, with pK_a values ranging from −13.4 to −19.4. Again, the calculated carbonyl pK_a goes from −18.1 to 2.0 when SMD solvation and an explicit water molecule are used. The hydroxyl moiety of 1c is not in resonance with the furan ring, and therefore inductively removes electron density from the ring system toward C5, whose calculated pK_a is −13.4 versus −14.7 in 1a. This results in the more negative pK_a values for C2 through C4. The furan oxygen is also calculated to be more basic by 1.2 pK_a units. Initial attempts to calculate a pK_a value for the hydroxyl group of 1c were unsuccessful because protonation of the hydroxyl oxygen of 1c yields an unstable structure, where the protonated hydroxyl group migrates to the C4 position during gas phase CBS-QB3 geometry optimization.¹⁹ Addition of an explicit water molecule hydrogen bonded to the alcohol group with SMD solvation eliminates this problem, however, and a pK_a of 2.6 is calculated for the hydroxyl group. Both the carbonyl and hydroxyl oxygens of 1c are thus the most basic³³ by a large margin, which is consistent with the dehydration shown in Scheme 5.

Reduction of the carbonyl of 1a to yield 1d makes the furyl ring more basic by 0.3–2.1 pK_a units (Scheme 9). Initial calculations of the pK_a of the alcohol group of 1d gave a value of −18.8. This again is very different from the relevant experimental value of −2 for 2-butanol³⁴ Addition of just one explicit water molecule with SMD solvation gives a calculated value of 2.3. Therefore the most basic site of 1d is the hydroxyl group.

A similar trend is found upon reduction of the carbonyl in 1b to give 1e: the furyl ring becomes more basic by 0.3–1.7 pK_a units with the exception of C2 (less basic by 0.6 pK_a units). In comparison to 1a and 1b, the furan ring of 1e is less basic than that of 1d by 1.6–5.6 pK_a units with the exception of C5, which is more basic by 0.7 pK_a units. Again, the pK_a of the hydroxyl group was initially calculated as −19.0, but rises to 2.3 when an explicit water molecule with SMD solvation is added, and is the most basic site of the molecule.

The α,β unsaturated compounds possess significantly different pK_a values. The furan ring carbon atoms of 3a have calculated pK_a values ranging from −17.3 to −21.1, i.e. these are less basic than those of 1a by approximately 3–5 pK_a units. The furan ring oxygen atom is also less basic with a pK_a of −35.8, while the olefinic carbon center, α to the ketone group has a pK_a value of −13.0. A pK_a value for the carbon atom β to the ketone moiety could not be calculated, as the proton on this carbon migrates across the double bond during geometry optimization.¹⁹ Like 1a–e, the calculated pK_a of the carbonyl oxygen in 3a increases significantly from −10.5 to 4.6 on inclusion of an explicit water molecule with SMD solvation. This is 4.2 pK_a units more basic than the ketone oxygen atom of 1a and illustrates the magnitude of the resonance stabilization provided by the enone moiety that is not available in 1a.

As illustrated above, in going from 1a to 1b, replacement of the 3a furyl methyl group with a proton to give 3b significantly alters the calculated pK_a values. In the case of 3b, the furan oxygen and C2–C4 sites are ∼1–5 pK_a units less basic than in 3a, while C5 is 1.1 units more basic than in 3a. The olefinic carbon α to the ketone carbonyl is 4.2 units less basic than that of 3a, with a pK_a of −17.2. As in 3a, protonation at the β olefinic carbon resulted in migration across the double bond during geometry optimization, thus preventing calculation of the corresponding pK_a value. Here again the ketone oxygen is the most basic site in 3b with a pK_a of −12.3 that increases to 3.6 on inclusion of an explicit water molecule with SMD solvation, (i.e. 3.5 units more basic than in 1b). That the ketone is the most basic site of 3a and 3b is consistent with an ion cyclotron resonance study of protonation of the structurally similar α,β unsaturated ketones 3-penten-2-one and 3-methyl-3-buten-2-one, in which the ketone-protonated analogues were found to be the most stable.³⁵

As in 1c, the inductive effect of the electron withdrawing hydroxyl group within 3c makes the C2–C4 carbon atoms of the furan ring comparatively less basic (with pK_a values ranging from −19.5 to −22.9). The C5 carbon atom is again more basic by ∼1 pK_a unit (−16.2) while the furan oxygen pK_a is identical to that of 3a (−35.8). The carbon atom α to the ketone group is less basic than that in 3a (pK_a of −14.2), and protonation at the β position once again results in migration across the double bond during optimization. More importantly, as in 1c, protonation at the 3c hydroxyl group yields an unstable structure that undergoes water molecule migration during gas phase CBS-QB3 geometry optimization.¹⁹ Again, as in the case of 1c, addition of an explicit water molecule hydrogen bonded to the hydroxyl group gives stable structures, and enables calculation of a pK_a of −6.6 for the hydroxyl group of 3c.¹⁹ This value increases to 0.2 when SMD solvation is used. Thus, as for 1c, the hydroxyl group is the most basic site of 3c. The ketone oxygen is the next most basic atom within 3c, with a calculated pK_a of −12.5 with an explicit water molecule added which increases to 3.6 with SMD solvation.

Discussion

Heating 1a simply with 10 mol% hydrochloric acid as catalyst at only 80–100 °C, in air leads to furan ring opening and a single product, 2a (Scheme 2). Under these mild conditions compound 1d is ring opened cleanly to 2d (Scheme 6) while 1b and 1c yield more complicated product mixtures (Schemes 3,4 and 5), as discussed below. Strong Brønsted acids work well as catalysts, while weaker Brønsted acids are much less effective. Studies of the ring opening of 1a in buffered solution indicate a specific acid catalysis mechanism, and this mechanism is also likely where ring opening is observed in other substrates tested here (e.g.1d).

The reaction solvent system also has a significant effect on the yield of the reaction, as evidenced by the test of various solvent systems using 1a shown in Table 1. The solubility of 1a was greatly improved in all of the mixtures of water and water-miscible organic solvent tested. This improved solubility did not necessarily lead to more product, however: pure water and a 1 : 1 mixture of water and methanol gave equivalent yields (86% vs. 87%) of 2a over 24 hours. Acetonitrile : water mixtures worked almost as well, giving 2a in 77% yield. In contrast, acetone, ethanol and isopropanol all gave significantly lower yields of 2a. Addition of an organic co-solvent is known to cause changes in ion dissociation and substrate pK_a values³⁶ as well as leading to preferential solvation.³⁷ The concentration of reactant water is also decreased, which may slow the reaction.

Some recent computational studies have examined various chemical transformations of biomass derivatives to chemical feedstocks,³⁸ but none have examined the ring opening of furans within these classes of compounds in detail.³⁹ As illustrated in Table 3, DFT calculations carried out at the 6-311+G(d) level indicate that ring opening of 1a–e to the corresponding products 2a, 2b*, 2c–e is downhill by 2.03 to 4.92 kcal mol⁻¹. The CBS-QB3 results indicate that ring opening is slightly endergonic, by 0.28 to 3.45 kcal mol⁻¹. In either case, the reaction will be further pushed towards product formation by the high concentrations of water present under the reaction conditions. The diverse reactivities of 1a−1e under similar conditions, however, (ring opening for 1a and 1d (Schemes 2 and 6), slight decomposition for 1b and 1e (Schemes 4 and 7), and decomposition to 3a for 1c) (Scheme 5) indicate that the kinetic pathways of these reactions are important in determining the observed products.

That the α,β unsaturated compounds (3a–3c) are unreactive toward ring opening (Scheme 8) is supported by both levels of theory: the overall reaction is calculated to be more uphill than in the case of their corresponding saturated counterparts (1a–c). For 3a and 3b, upon initial protonation at the carbonyl, both the rigidity of the butenone side chain due to enhanced π-conjugation and the poor basicities of the C2 positions should inhibit proton transfer to the furan ring and hence ring opening (Scheme 8).

The pK_a values calculated for 1a and 1b (Scheme 9) indicate that they are poor bases, and that the most basic site is the ketone oxygen. Because proton transfer does not occur in the rate limiting step, the most basic site in each substrate is not the sole influence in determining which products are observed during a specific acid catalysis mechanism. Consideration of pK_a values is instructive here however, as the pK_a values determine the pre-equilibrium species that form. Upon carbonyl protonation, the butanone side chain of 1a may be able to approach and protonate the furan C2, possibly through a water-assisted transition state,⁴⁰ and begin the observed ring opening mechanism. Alternatively, the species formed upon protonation of the ketone oxygen may be unreactive in terms of ring opening, and enough protonation at the furan ring occurs in solution for the reaction to occur. Indeed, protonation of the furyl C2 or C5 site is consistent with both experimental and theoretical work on furan and methyl furans.⁴¹ Either way, protonation of the ring of 1a could then be followed by nucleophilic attack of water leading to the ring opened triketone 2a; a theoretical study of the mechanism of this process is currently underway.

Furfural and HMF represent major components of biomass^4,5,42 and therefore are promising precursors for biofuels. Accordingly, compound 1b is synthesized using furfural while 1c is synthesized using HMF. In contrast to the reactions of 1a that yielded only one product (Scheme 2), reactions of 1b and 1c were more complex (Schemes 3, 4 and 5). Here, consideration of pK_a values may be more instructive, because for reactions where ring opening is not observed the rate limiting step is likely different. For 1b, the 19 carbon product 2b is obtained cleanly when heated in 1 : 1 water : acetone with HCl as a catalyst (Scheme 4). Compound 2b is the product of condensation of two molecules of 1b and one acetone molecule, and similar reactions have been reported using stoichiometric and catalytic amounts of acid as well as metal catalysts.⁴³ Slight decomposition of 1b is observed in other solvent mixtures, and no significant ring opening is observed. The furyl C2 position of 1b is comparatively less basic than in 1a (pK_a: −13.0 (1b) vs. −17.0 (1a)), which may prevent a water assisted proton transfer from the ketone oxygen to C2, thereby preventing ring opening.

The reactivity of 1c is also more complicated. When heated in the presence of acid catalyst, the major product is the α,β unsaturated compound 3a (Scheme 5), which forms via acid catalyzed dehydration. Calculation of an accurate pK_a value for the protonated alcohol moiety of 1c required inclusion of an explicit water molecule in addition to the solvent continuum. The resulting pK_a calculations show that the alcohol group is the most basic site of the molecule by 0.6 pK_a units.⁴⁴ The proximity of the alcohol group to the furan ring π-electrons thus apparently lowers the barrier for dehydration significantly. Furthermore, the thermodynamic calculations indicate that this dehydration reaction is quite downhill: going from 1c to 3a with release of H₂O has a ΔG^o of −15.0 kcal mol⁻¹ at the CPCM/B3LPY/6-311+G(d) level of theory and a ΔG^o of −10.2 kcal mol⁻¹ at the CPCM/CBS-QB3 level of theory. The α,β unsaturated compounds are unreactive under our experimental conditions, so this dehydration reaction substantially limits the yield of ring opened products. In terms of biofuels production, any 3a formed could be recovered, the olefin reduced to give 1a, which could then be ring opened. These purification and hydrogenation steps could add prohibitive costs however, and an alternative would be to dehydrate any HMF to methylfurfural⁴⁵ for synthesis of 1a. The other products formed in this reaction were not identified, but ¹H NMR spectra indicate that some ring opening does occur. These differences in products and selectivities among 1a–c under the same conditions highlight the sensitivity of the ring opening chemistry to the nature of the substrate.

That compound 1d ring opens cleanly to 2d (Scheme 6) while 1e decomposes slightly under the same conditions (Scheme 7) is interesting. The exact same trend is seen for 1a (Scheme 2) and 1b (Scheme 3), respectively. Furthermore, the computed pK_as with SMD solvation and explicit water molecules of the protonated carbonyl groups of 1a and 1b (0.4 and 0.1, respectively) are similar to those of the hydroxyl groups in 1d and 1e (both 2.3). Thus, the reaction pathways analogous to those postulated above for 1a and 1b should apply to substrates 1d and 1e as well. Despite the comparatively basic hydroxyl groups of 1d and 1e, no alcohol dehydration products are observed, in contrast to 1c. Acid catalyzed dehydration of secondary alcohols typically requires concentrated sulfuric or phosphoric acid and higher temperatures. In contrast to the dehydration of the primary alcohol in 1c (Scheme 5), the lack of dehydration of the secondary alcohols in 1d and 1e under the same conditions (Schemes 6 and 7) is likely due to the lack of π-electrons available to stabilize intermediates of dehydration of 1d or 1e.

Finally, calculated pK_a values proved very instructive in this study, and the success of these calculations, which are notoriously difficult,⁴⁶ rested on two things. First, calibration of the calculated values against the experimentally measured pK_a of 2,5 di-tert-butylfuran using various methods for solvation free energies ensured that accurate and thereby chemically intuitive pK_a values were obtained. Second, use of explicit hydrogen-bonded water molecules along with SMD solvation free energies proved critical for protonation of alcohol and ketone moieties. A few studies have shown that inclusion of one or more explicit solvent molecules can greatly increase the accuracy of pK_a calculations.⁴⁷ In the case of 1c and 3c, pK_a values for hydroxyl groups could not be calculated at all without inclusion of an explicit water molecule. For 1d and 1e, the calculated pK_a changed 11.0 and 10.6 units, respectively, upon addition of an explicit water molecule with SMD solvation. For the carbonyl groups of 1a–1c, use of an explicit water molecule improved the pK_as by 8.7–11.2 units.

Even better agreement of the calculated pK_as with experimental values may be possible by including additional explicit water molecules, but this would lead to a drastic increase in complexity due to the number of molecular conformations as water molecules are added.⁴⁸ These results highlight that while polarizable continuum solvent models may approximate the overall polarity of a given solvent, they cannot account for stabilization due to specific solvent interactions, especially hydrogen bonding. Hydrogen bond strengths are on the order of 4 kcal mol⁻¹,⁴⁹ and in aqueous acidic solution alcohol-water proton exchange can be significant.⁵⁰ A recent study found that inclusion of an explicit water molecule was not necessary for accurate calculation of pK_a values of cationic carbon acids if the charge on the carbon is sufficiently delocalized.⁴⁹ This is consistent with the results observed here: when calculating pK_a values for protonation of the furan ring of the substrates studied here, the furan π system helps delocalize the resulting positive charge and no explicit water molecules are required. No delocalization of charge is available to the alcohol moieties of 1d and 1e and carbonyl moieties of 1a–1c, making the effects of hydrogen bonding more important.

For protonated carbonyl and hydroxyl groups with explicit hydrogen bonded water molecules, replacing CPCM with SMD/HF/6-31G(d) solvation free energies in the thermochemical cycle improves the computed pK_as significantly. Indeed, the pK_as of the protonated alcohol groups in 1c, 1d, 1e, and 3c become 2.6, 2.3, 2.3, and 0.2, respectively, much closer to the experimental pK_a of −2 for 2-butanol.³³ In addition, the range pK_as for the protonated carbonyl groups of each substrate becomes 0.1–4.6, which is also in better agreement with the experimental value for acetone (−2 to −3).³¹ One of the factors responsible for the superior performance of SMD over CPCM in computing pK_as of protonated alcohol and carbonyl groups (in contrast to furan carbons) is likely the inclusion of such species (such as protonated acetone/acetophenone and protonated methanol/ethanol, each with an explicit hydrogen bonded water) in the SMD training set.^30,51

Conclusions

Furan rings contained within biomass-derived substrates can be ring opened under relatively mild conditions using simple, homogeneous Brønsted acid catalysts. Substituents on the furan rings can significantly alter the pathways of these proton catalyzed reactions. DFT calculations, aided by calibration to empirical values and inclusion of explicit water molecules when appropriate, support the notion that the location of initial protonation influences the products obtained. Compound 1a, containing a methyl group in the 4-position of the furan ring, gives one product in excellent yield via protonation of the furan ring and subsequent nucleophilic attack of a water molecule. When heated in an acetone:water mixture, compound 1b condenses with an acetone molecule and another molecule of 1b to give the nineteen carbon compound 2b. Only slight decomposition is observed when other solvent systems are used. The alcohol group within 1c is relatively basic, and prone to dehydration to yield the unreactive α,β unsaturated compound 3c and other unidentified products.

In contrast to the more typical high temperature, high pressure heterogeneous conditions used in the processing of biomass, these studies demonstrate that simple homogeneous catalysts can be effective under milder conditions and can also confer excellent selectivities. Despite much recent interest in biorenewables, there has been relatively little effort focused on understanding the intimate details of the chemistry required to produce fuels from biomass. This study improves the understanding of such processes, and may enable the design of better catalysts and processes for transformation of biomass to fuels and commodity chemicals on large scales.

Experimental section

Reagent grade chemicals and solvents were obtained from Aldrich, Acros or Fisher Scientific and used as received. ¹H and ¹³C NMR spectra were recorded on a Bruker Avance 400 MHz spectrometer at ambient temperatures and are referenced to a residual solvent peak. GC-MS analysis was obtained using a Hewlett-Packard 6890 GC system equipped with a Hewlett-Packard 5973 mass selective detector.

Compound 3a was obtained via a room temperature morpholine trifluoroacetate catalyzed aldol condensation of acetone and 5-methylfurfural,¹⁶ Similarly, compounds 3b and 3c were obtained via a room temperature, piperidine-catalyzed aldol condensation between acetone and furfural or hydroxymethylfurfural, respectively.¹⁶ Catalyst screening and ring opening reactions were performed in NMR tubes with 10 μL of substrate (.066 mmol) in 400 μL of solvent with 10 mol% catalyst added and the tubes were heated in a temperature controlled heating block. After heating, reaction mixtures were extracted three times with methylene chloride and the solvent removed in vacuo. The resulting residue was dissolved in CDCl₃, and a ¹H NMR spectrum was recorded with a delay time of 10 seconds to insure accurate integrals. Peak areas were calculated using the MestReNova line fitting tool.⁵²

Synthesis of 4-(5-methyl-2-furyl)-2-butanone (1a)

Compound 3a (5.95 g, 36.6 mmol) was reduced using magnesium in a methanol/THF mixture¹⁷ and purified by silica gel chromatography using 90 : 10 hexanes : Et₂O as eluent. Yields 4.64 grams (77%) ¹H NMR (CDCl₃): δ 2.16 (s, 3H, COCH₃), 2.23 (s, 3H, furan CH₃), 2.76 (t, 2H, CH₂), 2.85 (t, 2H, CH₂), 5.83 (d, 1H, CH), 5.85 (d, 1H, CH). ¹³C NMR (CDCl₃): δ 13.60, 22.43, 30.04, 42.07, 105.92, 106.05, 150.70, 152.77, 207.60.

4-(2-Furanyl)-2-butanone (1b)

Prepared analogously to 1a from 3b (0.206 g, 1.51 mmol) and purified by silica gel chromatography using methylene chloride as eluent. Yields 0.194 grams (93%) ¹H NMR (CDCl₃): δ 2.16 (s, 3H, CH₃), 2.78 (t, 2H, CH₂), 2.91 (t, 2H, CH₂), 5.99 (m, 1H, CH), 6.27 (m, 1H, CH), 7.29 (m, 1H, CH). ¹³C NMR (CDCl₃): 22.27, 29.37, 29.99, 41.78, 105.29, 110.31, 141.18, 207.39.

4-(5-Hydroxymethyl)-2-furanyl-2-butanone (1c)

Prepared analogously to 1a from 3c (0.198 g, 1.19 mmol) and purified by silica gel chromatography using Et₂O as eluent. Yields 0.138 grams (69%) ¹H NMR (CDCl₃): δ 2.17 (s, 3H, CH₃), 2.79 (t, 2H, CH₂), 2.90 (t, 2H, CH₂), 4.55 (s, 2H, CH₂OH), 5.94 (d, 1H, CH), 6.17 (d, 1H, CH). ¹³C NMR (CDCl₃): δ 22.37, 30.04, 41.75, 57.64, 106.18, 108.78, 152.71, 154.91, 207.41.

Synthesis of 4-(5-methylfuran-2-yl)butan-2-ol (1d)

A vial was filled with 5 mL of absolute ethanol and 50 μL of 1a (0.332 mmol). To this, 45 mg of NaBH₄ (1.18 mmol) was added and the solution was stirred for half an hour, followed by workup with dilute HCl, extraction with methylene chloride and removal of solvent in vacuo to yield 1d. Yields 0.047 grams (94%). ¹H NMR (CDCl₃): δ 1.22 (d, 3H, CHCH₃), 1.76 (m, 2H, CH₂), 2.25 (s, 3H, furan CH₃), 2.68 (m, 2H, CH₂), 3.84 (m, 1H, CH), 5.84 (m, 1H, CH), 5.86 (m, 1H, CH). ¹³C NMR (CDCl₃): δ 13.65, 23.64, 24.59, 37.69, 67.60, 105.65, 105.97, 150.54, 154.03.

Synthesis of 4-(furan-2-yl)butan-2-ol (1e)

Prepared analogously to 1d from 1b (0.0567 g, 0.41 mmol) and purified through a silica plug using methylene chloride as eluent. Yields 0.042 g (73%). ¹H NMR (CDCl₃): δ 1.23 (d, 3H, CHCH₃), 1.79 (m, 2H, CH₂), 2.74 (m, 2H, CH₂), 3.84 (m, 1H, CH), 6.00 (d, 1H, furan CH), 6.28 (d, 1H, furan CH), 7.3 (7.30, 1H, furan CH). ¹³C NMR (CDCl₃): δ 23.63, 24.48, 37.50, 67.46, 105.01, 110.26, 141.03, 155.91.

Synthesis of 2b

1b (0.010 g, 0.032 mmol) was heated at 80 °C for 24 hours in a 1 : 1 by volume mixture of water and acetone with 10 mol% HCl added. After heating, the reaction mixture was extracted three times with methylene chloride, dried over MgSO₄, and the solvent removed to give 2b in 92% yield. ¹H NMR (CDCl₃): δ 1.56 (s, 6H, CCH₃), 2.14 (s, 6H, COCH₃), 2.73 (t, 4H, CH₂), 2.86 (t, 4H, CH₂), 5.84 (m, 2H, CH), 5.86 (m, 2H, CH). ¹³C NMR (CDCl₃): δ 22.60, 26.45, 30.06, 37.44, 41.88, 104.61, 105.62, 153.01, 158.87, 207.64.

Acknowledgements

This work was supported by the Laboratory Directed Research and Development Program at LANL.

References

1. (a) J. C. Serrano-Ruiz and J. A. Dumesic, Energy Environ. Sci., 2011, 4, 83; (b) M. Stöcker, Angew. Chem., Int. Ed., 2008, 47, 9200–9211; (c) A. J. Ragauskas, C. K. Williams, B. H. Davison, G. Britovsek, J. Cairney, C. A. Eckert, W. J. Frederick, J. P. Hallet, D. J. Leak, C. L. Liotta, J. R. Mielenz, R. Murphy, R. Templer and T. Tschaplinski, Science, 2006, 311, 484–489; (d) G. W. Huber, S. Iborra and A. Corma, Chem. Rev., 2006, 106, 4044–4098.
2. Y. Zhang, in Biofuels from Agricultural Wastes and Byproducts, ed. H. P. Blaschek, T. Ezeji and J. Scheffran, Wiley-Blackwell, Hoboken, New Jersey, 2010, pp. 201–232.
3. (a) T. Barth and M. Kleinert, Chem. Eng. Technol., 2008, 31, 773–781; (b) T. R. Carlson, T. P. Vispute and G. W. Huber, ChemSusChem, 2008, 1, 397–400; (c) R. J. Evans and T. A. Milne, Energy Fuels, 1987, 1, 123–137; (d) L. Ingram, D. Mohan, M. Bricka, P. Steele, D. Strobel, D. Crocker, B. Mitchell, J. Mohammad, K. Cantrell and C. U. Pittman Jr., Energy Fuels, 2008, 22, 614–625; (e) S. Czernik and A. V. Bridgwater, Energy Fuels, 2004, 18, 590–598; (f) Z. Zhang, Q. Wang, P. Tripathi and C. U. Pittman Jr., Green Chem., 2011, 13, 940–949.
4. Y. Yu, X. Lou and H. Wu, Energy Fuels, 2008, 22, 46–60.
5. (a) Y. Roman-Leshkov, C. J. Barrett, Z. Y. Liu and J. A. Dumesic, Nature, 2007, 447, 982–985; (b) X. Tong, Y. Ma and Y. Li, Appl. Catal., A, 2010, 385, 1–13; (c) M. Mascal and E. B. Nikitin, ChemSusChem, 2009, 2, 423–426; (d) T. Thananatthanachon and T. B. Rauchfuss, Angew. Chem., Int. Ed., 2010, 122, 6766–6768.
6. R. M. West, Z. Y. Liu, M. Peter, C. A. Gartner and J. A. Dumesic, J. Mol. Catal. A: Chem., 2008, 296, 18–27.
7. G. W. Huber, J. N. Chheda, C. J. Barrett and J. A. Dumesic, Science, 2005, 308, 1446–1447.
8. M. Schlaf, Dalton Trans., 2006, 4545–4653.
9. For example, see: H. C. Lo, H. Han, L. J. D’Souza, S. C. Sinha and E. Keinan, J. Am. Chem. Soc., 2007, 129, 1246–1253 and references therein.
10. J. M. Watson, Ind. Eng. Chem. Prod. Res. Dev., 1973, 12, 310–311.
11. (a) S. Sayama, Synth. Commun., 2007, 37, 3067–3075; (b) W. W. Crew and R. J. Madix, J. Am. Chem. Soc., 1993, 115, 729–736; (c) O. Flogel and H. Reibig, Eur. J. Org. Chem., 2004, 2797–2804; (d) C. Dominguez, A. G. Csaky and J. Plumet, Tetrahedron Lett., 1990, 31, 7669–7670; (e) A. Mass, M. R. Acocella, M. De Rosa, A. Soriente, R. Villano and A. Scettri, Tetrahedron Lett., 2003, 44, 835–837; (f) J. Finlay, M. A. McKervey and H. Q. N. Gunaratne, Tetrahedron Lett., 1998, 39, 5651–5654; (g) J. Jurczak and S. Pikul, Tetrahedron Lett., 1985, 26, 3039–3040; (h) P. D. Williams and E. LeGoff, J. Org. Chem., 1981, 46, 4143–4147; (i) G. Piancatelli, A. Scettri and M. D’Auria, Tetrahedron, 1979, 36, 661–663; (j) S. H. Mousavipour, S. Ramazani and Z. Shahkolahi, J. Phys. Chem. A, 2009, 113, 2838–2846; (k) S. J. Blank and C. E. Stephens, Tetrahedron Lett., 2006, 47, 6849–6850; (l) L. Lepage and Y. Lepage, Synthesis, 1983, 1018–1019; (m) N. T. Do and M. Baerns, Appl. Catal., 1988, 9–26.
12. (a) J. Y. Gui, D. A. Stern, F. Lu and A. T. Hubbard, J. Electroanal. Chem., 1991, 305, 37–55; (b) S. Torii, H. Tanaka, Y. Tsutsui, T. Ohshima and S. Yamashita, Chem. Lett., 1986, 1535–1536.
13. J. Horvat, B. Klaic, B. Metelko and V. Sunjic, Tetrahedron Lett., 1985, 17, 2111–2114.
14. (a) A. Kankaanpera and P. Salomaa, Acta Chem. Scand., 1967, 21, 575–577; (b) A. Kankaanpera and S. Kleemola, Acta Chem. Scand., 1969, 23, 3607–3608.
15. (a) “Compounds and Methods For the Production of Long Chain Hydrocarbons From Biological Sources”, US Provisional Patent Application No. 61/534,496, filed September 14, 2011; (b) A. D. Sutton, F. Waldie, M. Schlaf, R. Wu, L. A. P Silks, J. C. Gordon, unpublished results.
16. J. M. Keith, J. K. Kim, L. Alexander, R. Wu, R. L. Martin, E. Batista, R. Michalczyk, B. L. Scott, A. D. Sutton, L. A. Silks and J. C. Gordon, manuscript submitted.
17. C. Dominguez, A. G. Csaky and J. Plumet, Tetrahedron Lett., 1991, 32, 4183–4184.
18. R. W. Alder, N. P. Hyland, J. C. Jeffrey, T. Riis-Johanessen and D. J. Riley, Org. Biomol. Chem., 2009, 7, 2704–2715.
19. In Los Alamos, with an elevation of approximately 7300 feet above sea level, water boils at ca. 93 °C; some mild refluxing was observed in the NMR tubes.
20. See ESI†.
21. P. K. Glasoe and F. A. Long, J. Phys. Chem., 1960, 64, 188–190.
22. T. Montagnon, D. Noutsias, I. Alexopoulou, M. Tofi and G. Vassilikogiannakis, Org. Biomol. Chem., 2011, 9, 2031–2039.
23. After 24 hours, starting material is still the predominant species in solution (93 to 97%). The solution darkening is suggestive of polymer formation, while some retro-aldol decomposition is also likely.
24. (a) A. D. Becke, J. Chem. Phys., 1993, 98, 1372–1378; (b) A. D. Becke, J. Chem. Phys., 1993, 98, 5648–5652.
25. (a) J. A. Montgomery Jr., M. J. Frisch, J. W. Ochterski and G. A. Petersson, J. Chem. Phys., 1999, 110, 2822–2827; (b) J. A. Montgomery Jr., M. J. Frisch, J. W. Ochterski and G. A. Petersson, J. Chem. Phys., 2000, 112, 6532–6542.
26. (a) V. Barone and M. Cossi, J. Phys. Chem. A, 1998, 102, 1995–2001; (b) M. Cossi, N. Rega, G. Scalmani and V. Barone, J. Comput. Chem., 2003, 24, 669–681.
27. (a) A. E. Reed and F. Weinhold, J. Chem. Phys., 1983, 78, 4066–4073; (b) A. E. Reed, R. B. Weinstock and F. Weinhold, J. Chem. Phys., 1985, 83, 735–746.
28. M. J. Frisch, et al., Gaussian 09, Revision B.01, Gaussian, Inc., Wallingford CT, 2009.
29. M. D. Liptak, K. C. Gross, P. G. Seybold, S. Feldgus and G. C. Shields, J. Am. Chem. Soc., 2002, 124, 6421–6427.
30. M. P. Carmody, M. J. Cook, N. L. Dassanayake, A. R. Katritzky, P. Linda and R. D. Tack, Tetrahedron, 1976, 32, 1767–1771.
31. A. V. Marenich, C. J. Cramer and D. G. Truhlar, J. Phys. Chem. B, 2009, 113, 6378–6396.
32. R. A. McLelland and W. F. Reynolds, Can. J. Chem., 1976, 54, 718–725.
33. For 1c, because the basicities of the hydroxyl and carbonyl groups are now similar, the most basic site could alternate between the two groups as additional explicit water molecules are added. For now, we tentatively assign the carbonyl as the most basic site of 1c.
34. N. C. Deno and J. O. Turner, J. Org. Chem., 1966, 31, 1969–1970 . For 2-butanol, we calculate pK_as of −18.3 and 9.4 with zero and one explicit water molecules, respectively. The latter value improves to 2.8 when SMD solvation is used. The same trend was found for acetone.
35. G. Bouchoux, Y. Hoppilliard, P. Jaudon and R. Houriet, Org. Mass Spectrom., 1984, 19, 394–397.
36. S. Sanli, Y. Altun, N. Sanli, G. Alsancak and J. L. Beltran, J. Chem. Eng. Data, 2009, 54, 3014–3021.
37. M. A. Silva, D. C. Silva, V. G. Machado, E. Longhinotti and V. L. A. Frescura, J. Phys. Chem. A., 2002, 106, 8820–8826.
38. (a) R. S. Assary, P. C. Assary, J. Redfern, J. Greeley and L. A. Curtiss, J. Phys. Chem. B, 2011, 115, 4341–4349; (b) R. S. Assary, P. C. Redfern, J. R. Hammond, J. Greeley and L. A. Curtiss, J. Phys. Chem. B, 2010, 114, 9002–9009; (c) T. Ståhlberg, S. Rodriguez-Rodriguez, P. Fristrup and A. Riisager, Chem.–Eur. J., 2011, 17, 1456–1464; (d) M. Zhang, Z. Geng and Y. Yu, Energy Fuels, 2011, 25, 2664–2670; (e) J. Guan, Q. Cao, X. Guo and X. Mu, Comput. Theor. Chem., 2011, 963, 453–462.
39. A similar transformation is studied in R. S. Assary, P. C. Redfern, J. R. Hammond, J. Greeley and L. A. Curtiss, Chem. Phys. Lett., 2010, 497, 123–128 . However, the authors added an additional step of furan dearomatization to THF prior to ring opening.
40. S. Caratzoulas and D. G. Vlachos, Carbohydr. Res., 2011, 346, 664–672 . In this QM/MM MD study of fructose conversion to 5-HMF, the authors report that a water assisted proton transfer from a protonated carbonyl to a hydroxyl in an intermediate step lowered the free energy barrier by 13.3 kcal mol⁻¹ relative to a non-assisted mechanism.
41. (a) R. Houriet, E. Rolli, G. Bouchoux and Y. Hoppilliard, Helv. Chim. Acta, 1985, 68, 2037–2045; (b) E. S. E. van Beelen, T. A. Koblenz, S. Ingemann and S. Hammerum, J. Phys. Chem. A, 2004, 108, 2787–2793; (c) K. Zeng and Z. Cao, Chin. J. Chem., 2006, 24, 293–298; (d) U. J. Lorenz, J. Lemaire, P. Maitre, M. E. Crestoni, S. Fornarini and O. Dopfer, Int. J. Mass. Spectrom., 2007, 267, 43–53.
42. H. Zhao, J. E. Holladay, H. Brown and Z. C. Zhang, Science, 2007, 316, 1597–1600.
43. (a) M. Skouta, A. Lesimple, Y. L. Bigot and M. Delmas, Synth. Commun., 1994, 24, 2571–2576; (b) R. Pajewski, R. Ostaszewski, K. Ziach, A. Kulesza and J. Jurczak, Synthesis, 2004, 6, 865–868; (c) A. S. K. Hashmi, L. Schwarz, P. Rubenbauer and M. C. Blanco, Adv. Synth. Catal., 2006, 348, 705–708.
44. For 1c, because the basicities of the hydroxyl and carbonyl groups are similar, the most basic site could alternate between the two groups as additional explicit water molecules are added. For now, we tentatively assign the hydroxyl as the most basic site of 1c. This caveat also applies to 3a, viz. the carbonyl group and the olefinic carbon α to the ketone.
45. G. Papadogianakis, L. Maat and R. A. Sheldon, J. Chem. Soc., Chem. Commun., 1994, 2659–2660.
46. For a discussion of some of the challenges involved in thermodynamic cycle-based pK_a calculations, see V. Bryantsev, M. Diallo and W. Goddard III, J. Phys. Chem. B, 2008, 112, 9709–9719.
47. (a) J. R. Pliego and J. M. Riveros, J. Phys. Chem. A, 2002, 106, 7434–7439; (b) C. P. Kelly, C. J. Cramer and D. G. Truhlar, J. Phys. Chem. A, 2006, 110, 2493–2499; (c) J. Ho and M. L. Coote, J. Chem. Theory Comput., 2009, 5, 295–306; (d) F. Eckert, M. Diedenhofen and A. Klamt, Mol. Phys., 2010, 108, 229–241.
48. See T. Miyake and M. Aida, Internet Electron. J. Mol. Des., 2003, 2, 24–32 for a discussion of the water trimer, tetramer and pentamer.
49. O. Markovitch and N. Agmon, J. Phys. Chem. A., 2007, 111, 2253–2256.
50. For t-BuOH, the rate constant for alcohol-water proton exchange is only 30% smaller than that for pure water. See E. Ralph and E. Grunwald, J. Am. Chem. Soc., 1969, 91, 2426–2429.
51. C. P. Kelly, C. J. Cramer and D. G. Truhlar, J. Chem. Theory Comput., 2005, 1, 1133–1152.
52. MestReNova, version 6.0.4-5850; 2010 Mestrelab Research S.L.

---

Footnote

† Electronic supplementary information (ESI) available: Complete citation for ref. 28, details of selected geometry optimizations, proton migrations, solvation free energies, and pK_as with CPCM/B3LYP/6-311+G(d) solvation. See DOI: 10.1039/c2cy20395b

---